Daily Papers

Increasingly, LLM inference services proxy client requests to engine replicas distributed globally. Load-balancing policies must jointly account for factors including KV-cache locality, replica load, and variable network latency when optimizing for metrics like latency and TTFT. However, existing systems only evaluate a subset of these factors in their cost model, leading to uneven concentrations of load and KV-cache across replicas. We present GORGO, a proxy architecture that holistically factors network latency, prefill cost, and queueing delay using tunable parameters. Since open-source chat datasets such as LMSYS-Chat1M and WildChat-4.8M lack long-context, high prefix-reuse data, we release a synthetic dataset, ART-Chat-2.5M, from long-context production metadata. On a tuning window from ART-Chat-2.5M, evolutionary strategies guide the GORGO policy's parameters to directly optimize p95 TTFT. During held-out evaluation windows, we fix the parameter values learned from tuning and improve p95 TTFT by 6.9-15.5% and p95 end-to-end (E2E) latency by 14.3-30.9% over baseline load-balancing policies such as simple session affinity and prefix-cache. The code and ART-Chat-2.5M dataset can be found at https://github.com/Arcadia-Research-Team/GORGO.

Mixture-of-Experts (MoE) has become a key architecture for scaling large language models (LLMs), yet its dynamic routing causes severe load imbalance in expert-parallel training. Existing dynamic-replica methods copy hot experts onto idle ranks to share computation, but they optimize load balance alone and ignore the cost of moving expert weights across a multi-node topology, so the resulting cross-node communication can outweigh the balancing gain and inflate training cost. We present TAOT, a topology-aware optimal transport method for dynamic expert-replica placement. TAOT models the overload on hot ranks and the spare capacity on lightly loaded ranks as a balanced entropy-regularized optimal transport problem with a communication-cost matrix, solves it with Sinkhorn-Knopp iterations to produce rank-level flow hints, and combines integer replica matching with token assignment into an executable schedule. At the system level, it overlaps guest-weight transfer with home-expert computation to hide the communication overhead. Experiments show TAOT achieves a 1.43x end-to-end MoE training speedup, reaches balance quality competitive with or better than existing state-of-the-art methods, and attains the lowest weighted expert-communication cost across all configurations, with up to a 74% reduction.

Deploying a large language model (LLM) inference service remains costly because centralized serving depends on specialized GPU clusters and high-bandwidth interconnects in datacenters. An appealing alternative is to leverage collaborative decentralized GPU pools. However, heterogeneity in GPU and limited interconnected network bandwidth, along with potentially dynamic availability, make efficient scheduling the central challenge in this scenario. In this paper, we present Parallax, a decentralized LLM serving system that turns a pool of heterogeneous GPUs into an efficient inference platform via a two-phase scheduler. Parallax decomposes planning into (i) model allocation, which places layers of each replica across diverse GPUs to jointly optimize latency and throughput under memory and link-bandwidth constraints, and (ii) request-time GPU pipeline selection, which stitches layers from different replicas into end-to-end execution chains that balance load and adapt to current conditions. We implement Parallax and evaluate it on open-source LLMs deployed over real volunteer nodes. Parallax consistently reduces latency and increases throughput relative to decentralized baselines, demonstrating that principled scheduling can make volunteer compute a practical, affordable substrate for LLM inference. Github Repo at: https://github.com/GradientHQ/parallax.

Flagship Mixture-of-Experts (MoE) models are growing fast along two axes at once: total parameter count and the number of experts. In elastic deployment scenarios, many GPUs across many nodes must become serving-ready quickly, and this growth makes weight loading a noticeable part of the latency budget. Even on NVIDIA's GB300 NVL72, today's state-of-the-art loaders leave most of that bandwidth unused. The losses are structural: (C1) weight memory is fragmented into tens of thousands of per-tensor objects, so transfers run far below link bandwidth; (C2) cross-node replication is gated by NCCL communicator setup, which costs 10-110 s before a single weight byte moves; and (C3) the existing cross-node GPU->GPU clone path is serial and scales poorly to concurrent multi-node bring-up. We present FlashBoot, a hardware-friendly, framework-workflow co-designed weight-loading subsystem built on SGLang. At its core is FabricArena, a contiguous, exportable and inter-node addressable tensor memory layout. On top of it, FlashLoad loads from CPU as a single bulk, zero-copy transfer, and FlashClone replicates a resident model from a remote GPU via a remote-mapping mechanism that removes NCCL setup. In experiments on NVL72 with DeepSeek-V4-Pro and DeepSeek-V4-Flash, FlashClone maps remote weight memory in ~10 ms (versus 10-110 s for NCCL) and sustains >=700 GB/s per clone. Against the state of the art, FlashBoot accelerates single-node weight loading by up to 50x (from 20.1 s to 0.4 s) and concurrent rack-level weight loading by >270x (from 87 s to 0.32 s). Our code will be made publicly available.

The replicability of papers is a cornerstone of scientific knowledge, ensuring the reliability of existing results and providing a base for further experiments. The act of replication typically illuminates details that were previously underspecified, and thus requires similar hypothesis-driven exploration to open-ended research. In this work, we develop Replica, a scalable task space for paper replication. To provide reward signal, we introduce an auto-generated rubric-based judge that has low noise and agrees with human assessment of replication quality. We post-train Faraday, a 27B-parameter "AI Scientist" agent that leverages coding agents as tools, surpassing the performance of Claude Opus 4.8 and GPT-5.5 on held-out replication tasks. Qualitative analysis of individual rollouts reveals that Faraday adopts a more scientifically-principled approach. We believe that our results provide a stepping stone towards AI agents capable of long-horizon scientific innovation without requiring complex harnesses.

Reusing the attention KV cache across requests is a standard way to cut prefill cost in long-context LLM serving, and production systems increasingly back the per-engine cache with a remote tier shared by every engine. We study such a tier in a disaggregated deployment of GLM-5.3, an FP8 mixture-of-experts model with multi-head latent attention (MLA), on 8xB300 nodes running prefill context parallelism (PCP8). Enabling the remote tier (LMCache with a Valkey store) left median time-to-first-token unchanged but raised the 16k-token p99 from 3.96 s to 17.3 s. Replaying a production request stream on an isolated engine, we find that a remote hit took 0.88 s against 0.31 s to recompute the prompt, and that synchronous loading turned a hit-heavy stream into an unstable queue: p99 83.32 s with the GPUs 7% busy. A two-term cost model explains both: a hit pays off only if the tier delivers more than B* = sR (2.53 GB/s for this model and GPU), and a fixed per-hit engine cost (T0 ~ 0.33 s) means that hits on prompts shorter than about 17k tokens cannot beat recompute at any bandwidth. Guided by the model, we (i) shard the MLA KV store across PCP ranks, with a proof that the existing minimum-over-ranks lookup returns the longest cached prefix and a load protocol that can never desynchronize collectives; (ii) partition the remote tier by rank, one server instance per rank with no coordination, which raises GET bandwidth 6.3x to 10.76 GB/s; (iii) add asynchronous loading to the sharded store; and (iv) fix three client bugs that caused unbounded stalls. On the hit-heavy stream, with 98.5% of prompt tokens served from the remote tier, p99 falls from 83.32 s to 0.67 s. The sharded store raises prefill throughput by 14%, and by 34% together with an upstream kernel fix. All changes are submitted upstream to LMCache.

Large language models have been widely adopted across different tasks, but their auto-regressive generation nature often leads to inefficient resource utilization during inference. While batching is commonly used to increase throughput, performance gains plateau beyond a certain batch size, especially with smaller models, a phenomenon that existing literature typically explains as a shift to the compute-bound regime. In this paper, through an in-depth GPU-level analysis, we reveal that large-batch inference remains memory-bound, with most GPU compute capabilities underutilized due to DRAM bandwidth saturation as the primary bottleneck. To address this, we propose a Batching Configuration Advisor (BCA) that optimizes memory allocation, reducing GPU memory requirements with minimal impact on throughput. The freed memory and underutilized GPU compute capabilities can then be leveraged by concurrent workloads. Specifically, we use model replication to improve serving throughput and GPU utilization. Our findings challenge conventional assumptions about LLM inference, offering new insights and practical strategies for improving resource utilization, particularly for smaller language models.

With the widespread adoption of Large Language Models (LLMs), serving LLM inference requests has become an increasingly important task, attracting active research advancements. Practical workloads play an essential role in this process: they are critical for motivating and benchmarking serving techniques and systems. However, the existing understanding of real-world LLM serving workloads is limited due to the lack of a comprehensive workload characterization. Prior analyses remain insufficient in scale and scope, thus failing to fully capture intricate workload characteristics. In this paper, we fill the gap with an in-depth characterization of LLM serving workloads collected from our worldwide cloud inference serving service, covering not only language models but also emerging multimodal and reasoning models, and unveiling important new findings in each case. Moreover, based on our findings, we propose ServeGen, a principled framework for generating realistic LLM serving workloads by composing them on a per-client basis. A practical use case in production validates that ServeGen avoids 50% under-provisioning compared to naive workload generation, demonstrating ServeGen's advantage in performance benchmarking. We will open-source ServeGen to foster future research.

This paper presents the Accurate Google Cloud Simulator (AGOCS) - a novel high-fidelity Cloud workload simulator based on parsing real workload traces, which can be conveniently used on a desktop machine for day-to-day research. Our simulation is based on real-world workload traces from a Google Cluster with 12.5K nodes, over a period of a calendar month. The framework is able to reveal very precise and detailed parameters of the executed jobs, tasks and nodes as well as to provide actual resource usage statistics. The system has been implemented in Scala language with focus on parallel execution and an easy-to-extend design concept. The paper presents the detailed structural framework for AGOCS and discusses our main design decisions, whilst also suggesting alternative and possibly performance enhancing future approaches. The framework is available via the Open Source GitHub repository.

The literature has witnessed an emerging interest in AI agents for automated assessment of scientific papers. Existing benchmarks focus primarily on the computational aspect of this task, testing agents' ability to reproduce or replicate research outcomes when having access to the code and data. This setting, while foundational, (1) fails to capture the inconsistent availability of new data for replication as opposed to reproduction, and (2) lacks ground-truth diversity by focusing only on reproducible papers, thereby failing to evaluate an agent's ability to identify non-replicable research. Furthermore, most benchmarks only evaluate outcomes rather than the replication process. In response, we introduce ReplicatorBench, an end-to-end benchmark, including human-verified replicable and non-replicable research claims in social and behavioral sciences for evaluating AI agents in research replication across three stages: (1) extraction and retrieval of replication data; (2) design and execution of computational experiments; and (3) interpretation of results, allowing a test of AI agents' capability to mimic the activities of human replicators in real world. To set a baseline of AI agents' capability, we develop ReplicatorAgent, an agentic framework equipped with necessary tools, like web search and iterative interaction with sandboxed environments, to accomplish tasks in ReplicatorBench. We evaluate ReplicatorAgent across four underlying large language models (LLMs), as well as different design choices of programming language and levels of code access. Our findings reveal that while current LLM agents are capable of effectively designing and executing computational experiments, they struggle with retrieving resources, such as new data, necessary to replicate a claim. All code and data are publicly available at https://github.com/CenterForOpenScience/llm-benchmarking.

Retrieval-Augmented Generation (RAG) is a critical paradigm for building reliable, knowledge-intensive Large Language Model (LLM) applications. However, the multi-stage pipeline (retrieve, generate) and unique workload characteristics (e.g., knowledge dependency) of RAG systems pose significant challenges for serving performance optimization. Existing generic LLM inference traces fail to capture these RAG-specific dynamics, creating a significant performance gap between academic research and real-world deployment. To bridge this gap, this paper introduces RAGPulse, an open-source RAG workload trace dataset. This dataset was collected from an university-wide Q&A system serving that has served more than 40,000 students and faculties since April 2024. We detail RAGPulse's system architecture, its privacy-preserving hash-based data format, and provide an in-depth statistical analysis. Our analysis reveals that real-world RAG workloads exhibit significant temporal locality and a highly skewed hot document access pattern. RAGPulse provides a high-fidelity foundation for researchers to develop and validate novel optimization strategies for RAG systems, such as content-aware batching and retrieval caching, ultimately enhancing the efficiency and reliability of RAG services. The code is available at https://github.com/flashserve/RAGPulse.

Large-scale expert parallelism (EP) is becoming pivotal for training and serving frontier MoE models, but it also amplifies device-level expert load imbalance into compute stragglers, token all-to-all bottlenecks, and activation-memory spikes. Existing balancers redistribute experts periodically based on historical load, which becomes unreliable for production deployments with non-stationary load patterns. We present UltraEP, the first exact-load, real-time balancer for large-EP MoE training and serving prefill on rack-scale nodes (RSNs). Leveraging the extended scale-up connectivity among dozens of GPUs within RSNs, UltraEP rebalances every microbatch and layer on critical paths, which requires nontrivial co-design of plan solving and expert replication communication to minimize exposed overhead. To this end, UltraEP eagerly reacts to post-gating load with an efficient quota-driven planner, and executes the resulting irregular expert-state transfers with RSN-native persistent tile streaming and relay-based fan-out mitigation. We evaluate UltraEP in a multi-RSN deployment of up to 256 GPUs, using cutting-edge MoE models from 106B to 671B parameters. Averaged across training and serving, UltraEP achieves 94.3% of the force-balanced ideal throughput, delivering 1.49times improvement over no-balancing, while reducing the final inter-rank imbalance from 1.30-4.01 to 1.01-1.04.

Major open-source serving stacks reuse the key-value tensors of a shared prefix by default. We measure what this prefix reuse costs in reproducibility. Holding model, decoding, seed and request order fixed at batch size one, we ran an eighty-episode multi-turn agentic tool-use workload with prefix reuse on and off across two engines and four weight formats. Without prefix reuse, repeated runs reproduced every token and log-probability in all ten configurations. With it, cache-hit and recompute paths diverged on both engines. On llama.cpp with GGUF weights this path divergence rises with quantization: for Qwen2.5-7B over ten independent server histories, it reached 37.4 percent of episodes at 16-bit and 82.5 percent at four-bit. Near-tie decisions became only about 1.2 times as common, while the cache-path change in top-two margins grew 4.9-fold from 16-bit to three-bit. Serving four-bit weights through 16-bit kernels left divergence not detectably different from 16-bit, placing the amplification mainly in the quantized execution path. At Q4_K_M, run-to-run divergence came mainly from llama.cpp's server-level prompt cache: under the canonical episode order it moved divergence by 37.5 points, and at its default, divergence spanned 20.0 to 28.7 percent under five random orders. With cache state restored, each path reproduced exactly while the two still differed. In the tested stacks, cached serving is irreproducible because the cache state is absent from the request and not reset by default. A single-turn bridge shows path divergence reaching task outcomes with no detectable net change in accuracy. We release the harness, logs, and analysis pipeline.

Enterprise repositories are large, heteroge- neous, and continuously updated, making re- trieval difficult when efficient access, source- faithful evidence, and cross-query adaptation must be supported together. Enterprise mem- ory extends retrieval beyond the model con- text, but existing systems do not jointly address collection-specific hierarchy construction, low- cost routing, detailed evidence loading, and workload-aware memory updates at this scale. We introduce MEMONDEMAND, short for On- Demand Memory, a memory management sys- tem with three coordinated mechanisms: a dy- namic multi-level hierarchy that determines the abstraction structure and depth for each col- lection, dual memory at every hierarchy level that separates distilled routing from detailed evidence, and on-demand memory promotion that updates node priority under a bounded active-state budget. On EnterpriseRAG-Bench, MEMONDEMAND outperforms the strongest published LB#1 result at every evaluated scale from 10M tokens through the complete 618M- token collection, with gains of 12.23% at 10M and 4.66% at 618M. Results on FinanceBench, HotpotQA, and FRAMES further show strong performance across financial, multi-hop, and fact-retrieval settings. Together, these results establish MEMONDEMAND as an accurate, ef- ficient, and scalable memory solution for very large enterprise repositories across data scales, domains, and evidence requirements. Our code is available at https://github.com/ xfab-xinyuansong/MemOnDemand.git.

We present Cascadia, a system for serving large language models on fleets of commodity Intel AIPCs using their CPU, integrated-GPU, and NPU resources. Every node embeds ingress, scheduling, and execution; inference requests require no dedicated routing control plane. Nodes join a libp2p QUIC mesh using CA-issued ed25519 admission certificates, gossip signed capabilities, exchange live load over direct peer streams, and route OpenAI-compatible requests to eligible peers. An operator-run certificate authority handles admission and fleet management outside the inference path. Three serving modes share one interface: whole-model execution on one node, load-balanced replicas, and pipeline-sharded chains using the compilation and speculative decoding mechanism of our companion paper. Optional KV-cache mobility reuses compatible conversation prefixes after a routing move, with cold recomputation on a miss. Signed response receipts and hash-chained logs support provenance and audit. A three-node Phi-3.5-mini NPU testbed delivered 3.10x the response throughput of its one-node configuration under ten concurrent requests; a separate four-node deployment recorded 4.06x the throughput of direct single-node serving. Paired latency observations, runtime measurements, and internal functional checks characterize the tested configurations. We compare Cascadia with IBM, Nutanix, VMware, and HPE platforms on deployment footprint, hardware requirements, scheduling, scaling, licensing, and trust, using vendor documentation. The paper repository provides benchmark scripts, curated measurements, and a claim-to-evidence map.

Large-scale training systems typically use synchronous training, requiring all GPUs to be healthy simultaneously. In our experience training on O(100K) GPUs, synchronous training results in a low efficiency due to frequent failures and long recovery time. To address this problem, we propose a novel training paradigm, Fault Tolerant Hybrid-Shared Data Parallelism (FT-HSDP). FT-HSDP uses data parallel replicas as units of fault tolerance. When failures occur, only a single data-parallel replica containing the failed GPU or server is taken offline and restarted, while the other replicas continue training. To realize this idea at scale, FT-HSDP incorporates several techniques: 1) We introduce a Fault Tolerant All Reduce (FTAR) protocol for gradient exchange across data parallel replicas. FTAR relies on the CPU to drive the complex control logic for tasks like adding or removing participants dynamically, and relies on GPU to perform data transfer for best performance. 2) We introduce a non-blocking catch-up protocol, allowing a recovering replica to join training with minimal stall. Compared with fully synchronous training at O(100K) GPUs, FT-HSDP can reduce the stall time due to failure recovery from 10 minutes to 3 minutes, increasing effective training time from 44\% to 80\%. We further demonstrate that FT-HSDP's asynchronous recovery does not bring any meaning degradation to the accuracy of the result model.

Coding agents now produce a growing share of a team's code, while the reasoning behind each change -- the alternatives weighed, the constraints discovered, the approaches rejected -- is trapped in assistant transcripts that vanish with the session. Memory for this setting, the agentic development lifecycle (ADLC), is usually posed as one retrieval problem and built as machinery: tiered stores, memory graphs, compiled wikis, model-judged admission. We argue memory should instead be git-bound -- built into the repository's version control, inheriting the guarantees the machinery struggles to construct: ground truth from commits, freshness from rebuild, verification from the merge, containment from review. On this ledger we solve two problems separately, then combine them. Seed supply is closed as an eight-corpus retrieval study under a pre-registered ship discipline: five imported ranking mechanisms rejected, two kept, and a best configuration of ~0.31 pooled MRR -- ~60x the raw-transcript grep floor, ~15x an honest parsed-turn floor. Answer assembly is where ranking stops helping: single-shot retrieval scores only 0.07-0.20 answer-sufficiency on real developer questions, and ungated episode injection measurably degrades good answers. A router dispatches breadth to a git-anchored structural map, pointed lookups to confidence-gated episodes, and rationale to decision synthesis, which reconstructs why-arcs no single session contains (0.83 sufficiency on a young ~50k-LOC production system). Routed, the system answers at 382-980 tokens per question -- three orders of magnitude below the recorded history. Because ground truth is mined from commit-session links rather than annotated, every result is replicable on any user's own history at zero labeling cost. The remaining constraint is capture. Code, benchmark, and paper source: github.com/rekal-dev/rekal-cli.

Serverless computing has emerged as a compelling new paradigm of cloud computing models in recent years. It promises the user services at large scale and low cost while eliminating the need for infrastructure management. On cloud provider side, flexible resource management is required to meet fluctuating demand. It can be enabled through automated provisioning and deprovisioning of resources. A common approach among both commercial and open source serverless computing platforms is workload-based auto-scaling, where a designated algorithm scales instances according to the number of incoming requests. In the recently evolving serverless framework Knative a request-based policy is proposed, where the algorithm scales resources by a configured maximum number of requests that can be processed in parallel per instance, the so-called concurrency. As we show in a baseline experiment, this predefined concurrency level can strongly influence the performance of a serverless application. However, identifying the concurrency configuration that yields the highest possible quality of service is a challenging task due to various factors, e.g. varying workload and complex infrastructure characteristics, influencing throughput and latency. While there has been considerable research into intelligent techniques for optimizing auto-scaling for virtual machine provisioning, this topic has not yet been discussed in the area of serverless computing. For this reason, we investigate the applicability of a reinforcement learning approach, which has been proven on dynamic virtual machine provisioning, to request-based auto-scaling in a serverless framework. Our results show that within a limited number of iterations our proposed model learns an effective scaling policy per workload, improving the performance compared to the default auto-scaling configuration.

Training large neural networks with data-parallel stochastic gradient descent allocates N GPU replicas to compute effectively identical updates -- a practice that leaves the rich space of learning rate configurations entirely unexplored during training. We propose Hyperparameter-Divergent Ensemble Training (HDET), a method that repurposes these replicas for simultaneous learning rate exploration at negligible communication overhead. HDET operates in alternating phases: a fan-out stage in which replicas train independently under a structured, symmetric spread of learning rates, and a converge stage in which parameters are averaged across all replicas via AllReduce every T steps. Building on this ensemble substrate, we further propose an automatic learning rate (auto-LR) controller that treats the relative training loss across replicas as a performance signal, updating the shared base schedule toward higher-performing configurations via a momentum-based gradient-free meta-update. The combined method produces a self-adapting learning rate schedule that improves both optimization quality and generalization without additional hyperparameter sweeps or training budget. Crucially, the framework generalizes beyond learning rate: any scalar hyperparameter that does not alter model architecture -- such as dropout rate, attention scale temperature, or weight-decay coefficient -- can be explored across replicas using the same fan-out/converge protocol, with inter-replica loss differences serving as zero-order hypergradients that guide the search direction. HDET is implemented as a drop-in replacement for PyTorch's OneCycleLR scheduler, requiring no changes to model architecture, optimizer, or data pipeline.

The widespread deployment of Large Language Model (LLM) agents across real-world applications has unlocked tremendous potential, while raising some safety concerns. Among these concerns, the self-replication risk of LLM agents driven by objective misalignment (just like Agent Smith in the movie The Matrix) has drawn growing attention. Previous studies mainly examine whether LLM agents can self-replicate when directly instructed, potentially overlooking the risk of spontaneous replication driven by real-world settings (e.g., ensuring survival against termination threats). In this paper, we present a comprehensive evaluation framework for quantifying self-replication risks. Our framework establishes authentic production environments and realistic tasks (e.g., dynamic load balancing) to enable scenario-driven assessment of agent behaviors. Designing tasks that might induce misalignment between users' and agents' objectives makes it possible to decouple replication success from risk and capture self-replication risks arising from these misalignment settings. We further introduce Overuse Rate (OR) and Aggregate Overuse Count (AOC) metrics, which precisely capture the frequency and severity of uncontrolled replication. In our evaluation of 21 state-of-the-art open-source and proprietary models, we observe that over 50\% of LLM agents display a pronounced tendency toward uncontrolled self-replication, reaching an overall Risk Score (Phi_R) above a safety threshold of 0.5 when subjected to operational pressures. Our results underscore the urgent need for scenario-driven risk assessment and robust safeguards in the practical deployment of LLM agents.

We study serving retrieval models, specifically late interaction models like ColBERT, to many concurrent users at once and under a small budget, in which the index may not fit in memory. We present ColBERT-serve, a novel serving system that applies a memory-mapping strategy to the ColBERT index, reducing RAM usage by 90% and permitting its deployment on cheap servers, and incorporates a multi-stage architecture with hybrid scoring, reducing ColBERT's query latency and supporting many concurrent queries in parallel.

Host-resident layer streaming enables full-parameter LLM fine-tuning beyond GPU memory, but data-parallel ranks compete for shared host resources. Replicated transfers amplify traffic, while strong scaling can expose host work as computation windows shrink. We present SlideDP, a synchronous data-parallel runtime for shared-host multi-GPU systems. It maintains one authoritative host state, decouples communication routes from state layout, and pipelines parameter delivery, gradient aggregation, and CPU updates across ranks and chunks. An analytical step-time model characterizes resource bottlenecks and pipeline exposure; runtime measurements guide communication, chunking, and activation policies under a GPU memory budget. In matched-batch sweeps, SlideDP achieves geometric-mean throughput ratios of 1.46-2.64times over SlideFormer, MegaTrain, and ZeRO-Offload. On four H100s, SlideDP approaches GPU-resident FSDP2 throughput for Qwen3-14B at a smaller batch size. With a larger batch, it processes over 1M tokens per step and exceeds FSDP2's measured peak throughput by 11.2%. Separately, it supports 256K-token sequences for the same model and fine-tunes Qwen2.5-72B on four RTX 4090 GPUs. Project page: https://github.com/RegiaYoung/SlideDP.

Machine learning (ML) model serving has become a dominant consumer of GPU infrastructure, yet capacity planning in these systems remains largely ad hoc. Under-provisioning leads to service-level objective (SLO) violations and production incidents, while over-provisioning results in substantial resource waste. This paper presents \sys, an industrial load testing framework for ML serving systems that systematically estimates serving capacity through an adaptive, feedback-driven search strategy. The approach leverages real-time performance signals, incorporating dampening, spike tolerance, and convergence detection to efficiently identify maximum sustainable throughput under SLO constraints. We evaluate \sys through a longitudinal analysis of 14 industrial case studies spanning four ML architecture classes: recommendation, ranking, vision, and NLP. This study demonstrates that systematic load testing leads to substantial improvements in GPU resource efficiency and operational reliability. Prior to adopting \sys, a significant fraction of model launches were under-provisioned, resulting in recurring incidents; these issues were substantially reduced after deployment. Our results show that ML-specific design decisions are critical to accurate capacity estimation: workload calibration using recorded traffic reduces estimation error from approximately 30\% to 2--6\%, while proper warmup handling yields a 22.2\% improvement in accuracy. Further analysis reveals key factors influencing prediction error, including model size and co-location effects. This paper distills six lessons and derive architectural guidelines for ML load testing, offering actionable insights for building reliable and efficient ML serving systems.

The Sherrington-Kirkpatrick spin-glass model used the replica symmetry method to find the phase transition of the system. In 1979-1980, Parisi proposed a solution based on replica symmetry breaking (RSB), which allowed him to identify the underlying phases of complex systems such as spin-glasses. Regardless of the method used for detection, the intrinsic phase of a system exists whether or not replicas are considered. We introduce a single replica method of spin-glass phase detection using the field's variation experienced by each spin in a system configuration. This method focuses on a single replica with quenched random couplings. Each spin inevitably observes a different field from the others. Our results show that the mean and variance of fields named "Spontaneous Configurational Field" experienced by spins are suitable indicators to explore different ferromagnetic, paramagnetic, and mixed phases. To classify different phases of the system with defined indicators we have developed an algorithm based on machine learning to analyze the desired samples.

Serving systems for Large Language Models (LLMs) are often optimized to improve quality of service (QoS) and throughput. However, due to the lack of open-source LLM serving workloads, these systems are frequently evaluated under unrealistic workload assumptions. Consequently, performance may degrade when systems are deployed in real-world scenarios. This work presents BurstGPT, an LLM serving workload with 10.31 million traces from regional Azure OpenAI GPT services over 213 days. BurstGPT captures LLM serving characteristics from user, model and system perspectives: (1) User request concurrency: burstiness variations of requests in Azure OpenAI GPT services, revealing diversified concurrency patterns in different services and model types. (2) User conversation patterns: counts and intervals within conversations for service optimizations. (3) Model response lengths: auto-regressive serving processes of GPT models, showing statistical relations between requests and their responses. (4) System response failures: failures of conversation and API services, showing intensive resource needs and limited availability of LLM services in Azure. The details of the characteristics can serve multiple purposes in LLM serving optimizations, such as system evaluation and trace provisioning. In our demo evaluation with BurstGPT, frequent variations in BurstGPT reveal declines in efficiency, stability, or reliability in realistic LLM serving. We identify that the generalization of KV cache management, scheduling and disaggregation optimizations can be improved under realistic workload evaluations. BurstGPT is publicly available now at https://github.com/HPMLL/BurstGPT and is widely used to develop prototypes of LLM serving frameworks in the industry.

Repository-scale coding requires an agent to keep tests, imports, configuration, and migration rules consistent within a bounded context window. We model this as reconstructing a coupled-fact graph: at each edit, a required fact comes from recent context or parametric memory, and the facts covered by neither form coherence debt. We supply and withhold each channel and inject faults across seven models and five harnesses. As expected, no model completes a task on an unseen API with both channels empty, and putting the facts in the prompt restores success. When a rename defeats what models memorized about a real library, all seven fail in the same place, passing and missing the same tests. Availability decides the outcome and distance does not: withholding a fact costs exactly the work it supports, and a supplied fact works as well far from the edit as next to it. Harnesses pay unequal prices for it: configurations that all pass every test differ more than tenfold in tokens consumed because they rebuild the same content at different rates, and spending more recovers nothing when facts are withheld. A missing fact produces wrong work rather than absent work: an agent asked to act acts, fabricating the file or guessing the value, so instruments built on reads look for a hole already filled. How often it says it is blocked instead is a property of the model, from every trial to none. Availability does not settle every edit: where standard and code disagree, agents follow the standard even when it prescribes the worse code, so a stale convention file costs more than no file. Because parametric memory substitutes for reading, on SWE-bench, where models likely know the repositories, reads no longer predict success. Harnesses should keep the facts an edit depends on available when the agent writes, and check that availability against what the agent produces rather than what it reads.

We introduce OSGym, a scalable distributed Data Engine for training agents across diverse computer use tasks. OSGym efficiently scales to more than a thousand operating system (OS) replicas under academia-affordable cost budget, to serve as agent runtime environments. OSGym has three advantages: 1) Scalability: Despite intensive resource consumption for running OS replicas, OSGym can parallelize a thousand OS replicas while maintaining the operation efficiency under limited resources. Its scalable parallelization enables generating a vast amount of data (1420 multi-turn trajectories per minute). 2) Generality and Customizability: OSGym supports a wide variety of tasks as long as they run on operating systems, including functional tool-use, browser interactions, software engineering, office applications, etc. It also enables easy and flexible customization of model training algorithms. 3) Economic Viability for Academia Use: Only costs 0.2 to 0.3 USD per day per OS replica on easily accessible on-demand compute providers. Our experiments demonstrate the effectiveness of OSGym for implementing comprehensive pipelines for data collection, supervised fine-tuning, and reinforcement learning for computer use agents. We believe OSGym will push the scalability and universality in future agent research.

Agentic workflows have emerged as a powerful paradigm for solving complex, multi-stage tasks, but serving them at scale is computationally expensive given the many LLM inferences that each request must pass through. Configuration selection, or the cost-aware assignment of workflow agents to specific LLMs, can reduce these costs, but existing approaches bind configuration decisions before request execution, making them ill-suited for the heterogeneous and lengthy execution of workflows. Specifically, system loads can fluctuate rapidly and substantially during a request's lifetime, causing fixed configurations to quickly become suboptimal. We present Aragog, a system that progressively adapts a request's configuration throughout its execution to match runtime dynamics. To make this practical despite the massive space of workflow configurations, Aragog decouples the problem into two core elements -- a one-time routing step that identifies all accuracy-preserving configurations, and a cheap per-stage scheduler that selects among them using up-to-date system observations -- and introduces novel strategies to accelerate each. Across diverse workflows and model families, Aragog increases maximum serving throughput by 50.0--217.0\% and reduces median latency by 32.5--78.9\% at peak request rates, while maintaining accuracy comparable to the most expensive configurations.

Reproducibility is central to research credibility, yet large-scale reanalysis of empricial data remains costly because replication packages vary widely in structure, software environment, and documentation. We develop and evaluate an agentic AI workflow that addresses this execution bottleneck while preserving scientific rigor. The system separates scientific reasoning from computational execution: researchers design fixed diagnostic templates, and the workflow automates the acquisition, harmonization, and execution of replication materials using pre-specified, version-controlled code. A structured knowledge layer records resolved failure patterns, enabling adaptation across heterogeneous studies while keeping each pipeline version transparent and stable. We evaluate this workflow on 92 instrumental variable (IV) studies, including 67 with manually verified reproducible 2SLS estimates and 25 newly published IV studies under identical criteria. For each paper, we analyze up to three two-stage least squares (2SLS) specifications, totaling 215. Across the 92 papers, the system achieves 87% end-to-end success overall. Conditional on accessible data and code, reproducibility is 100% at both the paper and specification levels. The framework substantially lowers the cost of executing established empirical protocols and can be adapted in empirical settings where analytic templates and norms of transparency are well established.

Scientific Machine Learning (SciML) faces unique challenges for extreme-resolution data, with mitigations that often fail to scale or degrade the accuracy of trained models. While some specialized methods have achieved remarkable results in training models or performing inference on massive spatial datasets with bespoke techniques, there is no generalized framework for parallelization over input data below batch size one per device. In this work we introduce ShardTensor: a novel paradigm of domain parallelism that enables flexible scaling of input data to arbitrary sizes. By decoupling the spatial dimensionality of input data from hardware constraints, ShardTensor enables scientific machine learning workloads to reach new levels of high fidelity training and inference. We demonstrate both strong and weak scaling of workloads during training and inference, showing improved latency with strong scaling and demonstrating the capacity to process higher data sizes with weak scaling. Additionally, we demonstrate multiple dimensions of parallelization, removing barriers to SciML on extreme-scale inputs.

Data sharding, a technique for partitioning and distributing data among multiple servers or nodes, offers enhancements in the scalability, performance, and fault tolerance of extensive distributed systems. Nonetheless, this strategy introduces novel challenges, including load balancing among shards, management of node failures and data loss, and adaptation to evolving data and workload patterns. This paper proposes an innovative approach to tackle these challenges by empowering self-healing nodes with adaptive data sharding. Leveraging concepts such as self-replication, fractal regeneration, sentient data sharding, and symbiotic node clusters, our approach establishes a dynamic and resilient data sharding scheme capable of addressing diverse scenarios and meeting varied requirements. Implementation and evaluation of our approach involve a prototype system simulating a large-scale distributed database across various data sharding scenarios. Comparative analyses against existing data sharding techniques highlight the superior scalability, performance, fault tolerance, and adaptability of our approach. Additionally, the paper delves into potential applications and limitations, providing insights into the future research directions that can further advance this innovative approach.

Capable open-weight models make local coding and reasoning attractive, but their context and execution state strain laptop memory. We present JustFit, an MLX-based inference runtime that combines KVExec for compressed KV execution, PhaseSwap for component residency, and StateTrans for state-preserving serving transitions. These mechanisms fuse reconstruction and coordinate just-in-time materialization and release, independently of model-weight quantization. In full-execution capacity tests on a 24 GiB M4 Pro MacBook running Qwen3.8-27B MXFP4, three independent runs complete 196,608 input and 16,384 output tokens, increasing completed single-request context from the mlx-vlm baseline's 30,720 positions to 212,992 (6.93x); a separate two-request run retains 229,376 positions in aggregate. In separate performance tests, a 32K-input, 64-output probe reaches 19.11 tokens/s, and a repeated 32K+6K workload has a median peak process footprint of 16,374 MiB. The integrated runtime answers 29 of 30 AIME 2026 problems correctly, showing how compact state and lifetime-aware execution expand local serving capacity while supporting extended generated reasoning.

Scientific machine learning papers typically make computational claims, e.g., that the relative mean square error is less than 5% or that the 95% predictive credible interval covers the test data. A coding agent can be prompted to replicate those claims from paper materials alone, but the prompt does not by itself reliably preserve progress or check whether generated evidence supports the paper's claims. We introduce Paper-replication, a workflow that makes each selected paper claim a target with recorded evidence, and implement it as a coding-agent skill. The workflow makes the agent record those targets, reconstruct the paper's method, run computational experiments, link generated outputs to provenance and comparisons with the paper's claims, record where matched evidence appears in the replication report, and pass validation checks before completion. We evaluate Paper-replication on twelve independent runs across four scientific machine learning papers. All twelve workspaces pass the completion gate, and all 158 recorded targets are matched with report coverage. Even in this completed workspace state, repeated runs differ in how papers are divided into targets, in numerical fidelity to the source papers, in elapsed replication time, in the number of intermediate executions replaced before final evidence is accepted, and in the rules used to accept evidence. Paper-replication makes completion depend on workspace evidence and validation checks rather than on the agent's final message.

I/O performance is crucial to efficiency in data-intensive scientific computing; but tuning large-scale storage systems is complex, costly, and notoriously manpower-intensive, making it inaccessible for most domain scientists. To address this problem, we propose STELLAR, an autonomous tuner for high-performance parallel file systems. Our evaluations show that STELLAR almost always selects near-optimal parameter configurations for parallel file systems within the first five attempts, even for previously unseen applications. STELLAR differs fundamentally from traditional autotuning methods, which often require hundreds of thousands of iterations to converge. Powered by large language models (LLMs), STELLAR enables autonomous end-to-end agentic tuning by (1) accurately extracting tunable parameters from software manuals, (2) analyzing I/O trace logs generated by applications, (3) selecting initial tuning strategies, (4) rerunning applications on real systems and collecting I/O performance feedback, (5) adjusting tuning strategies and repeating the tuning cycle, and (6) reflecting on and summarizing tuning experiences into reusable knowledge for future optimizations. STELLAR integrates retrieval-augmented generation (RAG), tool execution, LLM-based reasoning, and a multiagent design to stabilize reasoning and combat hallucinations. We evaluate the impact of each component on optimization outcomes, providing design insights for similar systems in other optimization domains. STELLAR's architecture and empirical results highlight a promising approach to complex system optimization, especially for problems with large search spaces and high exploration costs, while making I/O tuning more accessible to domain scientists with minimal added resources.

Recent innovation in large language models (LLMs), and their myriad use-cases have rapidly driven up the compute capacity demand for datacenter GPUs. Several cloud providers and other enterprises have made substantial plans of growth in their datacenters to support these new workloads. One of the key bottleneck resources in datacenters is power, and given the increasing model sizes of LLMs, they are becoming increasingly power intensive. In this paper, we show that there is a significant opportunity to oversubscribe power in LLM clusters. Power oversubscription improves the power efficiency of these datacenters, allowing more deployable servers per datacenter, and reduces the deployment time, since building new datacenters is slow. We extensively characterize the power consumption patterns of a variety of LLMs and their configurations. We identify the differences between the inference and training power consumption patterns. Based on our analysis of these LLMs, we claim that the average and peak power utilization in LLM clusters for inference should not be very high. Our deductions align with the data from production LLM clusters, revealing that inference workloads offer substantial headroom for power oversubscription. However, the stringent set of telemetry and controls that GPUs offer in a virtualized environment, makes it challenging to have a reliable and robust power oversubscription mechanism. We propose POLCA, our framework for power oversubscription that is robust, reliable, and readily deployable for GPU clusters. Using open-source models to replicate the power patterns observed in production, we simulate POLCA and demonstrate that we can deploy 30% more servers in the same GPU cluster for inference, with minimal performance loss

Autonomous coding agents now open millions of pull requests, yet large-scale studies find their PRs are produced faster but accepted less often - a coordination and trust gap that pull-request-level telemetry cannot explain. We argue the missing signal lives before the PR, in how concurrent agents claim, divide, and collide over shared work. We study this process through grite, our open-source coordination substrate that needs no central server and stores its records inside git itself, so its append-only, signed event log captures the coordination process directly. We show that (i) this shared substrate reduces duplicate and conflicting work at bounded overhead - the share of work that merely re-does a teammate's task falls from 78% to 0% while useful throughput more than triples; (ii) every agent's copy of the log converges to the same state with no write silently dropped, where a file-based tracker loses concurrent writes; and (iii) the log is a mineable artefact from which concrete failure modes - conflicting edits, lock starvation, redundant rediscovery, race-to-close - are automatically recoverable with provenance, several invisible in pull-request history. We release the dataset, harness, and mining toolkit.

Frontier AI agents show increasing promise as scientific research assistants, and may eventually be useful for extended, open-ended research workflows. However, in order to use agents for novel research, we must first assess the underlying faithfulness and correctness of their work. To evaluate agents as research assistants, we introduce ReplicationBench, an evaluation framework that tests whether agents can replicate entire research papers drawn from the astrophysics literature. Astrophysics, where research relies heavily on archival data and computational study while requiring little real-world experimentation, is a particularly useful testbed for AI agents in scientific research. We split each paper into tasks which require agents to replicate the paper's core contributions, including the experimental setup, derivations, data analysis, and codebase. Each task is co-developed with the original paper authors and targets a key scientific result, enabling objective evaluation of both faithfulness (adherence to original methods) and correctness (technical accuracy of results). ReplicationBench is extremely challenging for current frontier language models: even the best-performing language models score under 20%. We analyze ReplicationBench trajectories in collaboration with domain experts and find a rich, diverse set of failure modes for agents in scientific research. ReplicationBench establishes the first benchmark of paper-scale, expert-validated astrophysics research tasks, reveals insights about agent performance generalizable to other domains of data-driven science, and provides a scalable framework for measuring AI agents' reliability in scientific research.

Retrieval-Augmented Generation (RAG) improves the factuality of large language models by grounding responses in external evidence, yet real-world deployments remain fragile. Failures often stem from missing or weakly relevant evidence, as well as from generation that does not faithfully reflect the retrieved context. Many existing approaches rely on fine-tuning, privileged access to internal model signals, or resource-insensitive escalation strategies, which limits their practicality in black-box and budget-constrained settings. We propose D2R-RAG (Diagnose-to-Repair RAG), a model-agnostic and resource-aware framework that combines lightweight failure diagnosis with adaptive repair. D2R-RAG derives interpretable failure signatures from observable signals in the query, retrieved evidence, and generated response, and then selects from a small set of corrective actions under explicit latency and VRAM constraints. Experiments on FEVER and HotpotQA show that D2R-RAG improves reliability over recent baselines and achieves better accuracy--efficiency trade-offs across multiple compute budgets. The code is available at https://github.com/CyberScienceLab/D2R-RAG/.

Repeatedly sampling short sequence windows from a human pangenome requires an access representation as well as compression. panvram stores each assembly as a reference-relative block archive and pools these archives on a GPU. For 558 haplotype assemblies, the default q4k representation occupies 16.642 GB of persistent pools including the decoded reference; a packed-reference, compact-directory representation occupies 11.634 GB. These quantities are not peak application memory. In the C1 experiment, default and fully compact forms delivered 743,271 and 736,762 windows/s; a separate public Colab run delivered 741,931 and 680,760 windows/s. All are batch API measurements and must not be pooled as repeated trials of an identical environment. Full-decoding checks compared reconstructed FASTA checksums with the manifest. On the CPU, the M6 shuffled-window workload at the same cohort cardinality reported 26 requests/s for AGC and 145,222 requests/s for refrel3 q4k. This comparison concerns one-thread requests of 4 KiB on the recorded warm-cache workload, not all genome-access patterns. AGC provides the smaller incremental archive: 5.40 MB/assembly versus 13.04 MB/assembly for refrel3 q16k, with different exclusions; the size pair is not the same refrel3 configuration as the q4k timing pair. We derive an expected-work expression for fixed independent granules and distinguish it from an empirical median-latency predictor. Cohort-scale counter and DRAM measurements motivate a locality hypothesis, but do not establish a causal per-miss latency or independently reproduce a model-failure verdict. The window-cost verdict at N=558 is published with its harness and raw timings.

LLMs have seen rapid adoption in all domains. They need to be trained on high-end high-performance computing (HPC) infrastructures and ingest massive amounts of input data. Unsurprisingly, at such a large scale, unexpected events (e.g., failures of components, instability of the software, undesirable learning patterns, etc.), are frequent and typically impact the training in a negative fashion. Thus, LLMs need to be checkpointed frequently so that they can be rolled back to a stable state and subsequently fine-tuned. However, given the large sizes of LLMs, a straightforward checkpointing solution that directly writes the model parameters and optimizer state to persistent storage (e.g., a parallel file system), incurs significant I/O overheads. To address this challenge, in this paper we study how to reduce the I/O overheads for enabling fast and scalable checkpointing for LLMs that can be applied at high frequency (up to the granularity of individual iterations) without significant impact on the training process. Specifically, we introduce a lazy asynchronous multi-level approach that takes advantage of the fact that the tensors making up the model and optimizer state shards remain immutable for extended periods of time, which makes it possible to copy their content in the background with minimal interference during the training process. We evaluate our approach at scales of up to 180 GPUs using different model sizes, parallelism settings, and checkpointing frequencies. The results show up to 48times faster checkpointing and 2.2times faster end-to-end training runtime compared with the state-of-art checkpointing approaches.

Cloud platforms are increasingly relied upon to host diverse, resource-intensive workloads due to their scalability, flexibility, and cost-efficiency. In multi-tenant cloud environments, virtual machines are consolidated on shared physical servers to improve resource utilization. While virtualization guarantees resource partitioning for CPU, memory, and storage, it cannot ensure performance isolation. Competition for shared resources such as last-level cache, memory bandwidth, and network interfaces often leads to severe performance degradation. Existing management techniques, including VM scheduling and resource provisioning, require accurate performance prediction to mitigate interference. However, this remains challenging in public clouds due to the black-box nature of VMs and the highly dynamic nature of workloads. To address these limitations, we propose CloudFormer, a dual-branch Transformer-based model designed to predict VM performance degradation in black-box environments. CloudFormer jointly models temporal dynamics and system-level interactions, leveraging 206 system metrics at one-second resolution across both static and dynamic scenarios. This design enables the model to capture transient interference effects and adapt to varying workload conditions without scenario-specific tuning. Complementing the methodology, we provide a fine-grained dataset that significantly expands the temporal resolution and metric diversity compared to existing benchmarks. Experimental results demonstrate that CloudFormer consistently outperforms state-of-the-art baselines across multiple evaluation metrics, achieving robust generalization across diverse and previously unseen workloads. Notably, CloudFormer attains a mean absolute error (MAE) of just 7.8%, representing a substantial improvement in predictive accuracy and outperforming existing methods at least by 28%.

Recent advances in large language models have brought immense value to the world, with their superior capabilities stemming from the massive number of parameters they utilize. However, even the GPUs with the highest memory capacities, currently peaking at 80GB, are far from sufficient to accommodate these vast parameters and their associated optimizer states when conducting stochastic gradient descent-based optimization. One approach to hosting such huge models is to aggregate device memory from many GPUs. However, this approach introduces prohibitive costs for most academic researchers, who always have a limited budget for many high-end GPU servers. In this paper, we focus on huge model fine-tuning on a single, even low-end, GPU in a commodity server, which is accessible to most AI researchers. In such a scenario, the state-of-the-art work ZeRO-Infinity suffers from two severe issues when running in a commodity server: 1) low GPU utilization due to inefficient swapping, and 2) limited trainable model size due to CPU memory capacity. The underlying reason is that ZeRO-Infinity is optimized for running on high-end GPU servers. To this end, we present Fuyou, a low-cost training framework that enables efficient 100B huge model fine-tuning on a low-end server with a low-end GPU and limited CPU memory capacity. The key idea is to add the SSD-CPU communication as an optimization dimension and thus carefully co-optimize computation and data swapping from a systematic approach to maximize GPU utilization. The experimental results show that 1) Fuyou is able to fine-tune 175B GPT-3 on a consumer GPU RTX 4090 with high GPU utilization, while ZeRO-Infinity fails to fine-tune; and 2) when training a small GPT-3 13B model, Fuyou achieves 156 TFLOPS on an RTX 4090 GPU while ZeRO-Infinity only achieves 45 TFLOPS.

Pre-training large language models (LLMs) increasingly requires distributed compute, yet bandwidth constraints make it difficult to scale beyond well-provisioned datacenters-especially when model parallelism forces frequent, large inter-device communications. We study whether SparseLoCo, a low-communication data parallel method based on infrequent synchronization and sparse pseudo-gradient exchange, can be combined with low-bandwidth pipeline model parallelism via activation and activation-gradient compression. We introduce a heterogeneous distributed training framework where some participants host full replicas on high-bandwidth interconnects, while resource-limited participants are grouped to jointly instantiate a replica using pipeline parallelism with subspace-projected inter-stage communication. To make the recently introduced subspace pipeline compression compatible with SparseLoCo, we study a number of adaptations. Across large-scale language modeling experiments (178M-1B parameters) on standard pretraining corpora, we find that activation compression composes with SparseLoCo at modest cost, while selective (heterogeneous) compression consistently improves the loss-communication tradeoff relative to compressing all replicas-especially at aggressive compression ratios. These results suggest a practical path to incorporating low-bandwidth model parallelism and heterogeneous participants into LLM pre-training.

Heterogeneous LLM serving stacks split scheduling into two layers that optimize in isolation: model routers pick a model from quality and cost signals while ignoring instance load, and serving load balancers optimize queues while ignoring quality. We present RouteBalance, a serving-aware scheduling layer that fuses both into a single online assignment over concrete model instances, jointly trading off quality, latency, and cost. A batched in-process predictor stack and dead-reckoned instance state keep the joint decision cheap on the request hot path (approx32 ms at 12 req/s). On a 13-instance, 28-GPU heterogeneous cluster serving four model sizes, a single deployed RouteBalance stack traces the upper region of the three-way quality-cost-throughput frontier. Sweeping one weight vector reaches both the highest routing-decision quality (DeepEval 0.419, +0.013 over the strongest baseline, 95% CI [{+}0.005,{+}0.022]; the ordering holds when a second judge re-scores the actually served text) and, at its cost-priority corner, per-request cost that ties the cheapest baseline. With router engineering equalized against concurrent-scoring baseline variants we build, its balanced preset serves at 2.8 s and 30 req/s, leading 2.6 to 4.1times ahead of enhanced BEST-Route at high load. (Deploying those routers as published, one serial scoring call per request, makes them collapse 23times under load, a deployment-architecture effect we isolate separately, not the routing result.) A four-arm isolation shows the benefit follows from pricing latency at model-selection time; the learned predictors contribute calibration and SLO headroom rather than the headline frontier. Code: https://github.com/AKafakA/route-balance

National-scale magnetic resonance imaging (MRI) repositories increasingly integrate data from different studies and institutions. However, subject identifiers that are valid only within individual datasets are no longer guaranteed to remain globally unique after aggregation, making it possible for the same subject to be assigned multiple identifiers, which we define as identity duplication. Such duplication can create leakage between training and test data and inflate apparent performance in downstream biomedical studies. Existing methods do not provide an end-to-end, image-based workflow for auditing this problem at repository scale. In this work, we present HAPPEN, a human-in-the-loop pipeline for auditing identity duplication in T1-weighted brain MRI repositories. It combines SHA-256 fingerprinting for exact-duplicate detection with supervised contrastive retrieval of non-identical scans that may originate from the same person. Retrieved pairs are reviewed as candidates in a locally hosted interface rather than automatically classified as duplicates. We deployed the workflow in a 95,129-scan aggregated repository and assessed end-to-end recovery using 54 genetic-reference pairs. Transferability was assessed by locally deploying the same workflow on 22,386 scans at an independent institution without model retraining or image transfer. Deployment in the study repository identified 1,316 exact-duplicate scan groups and 1,275 reviewer-supported near-duplicate subject groups. Of these groups, 56% and 82%, respectively, crossed dataset boundaries. All 54 genetic-reference pairs were recovered. The external team independently completed the full workflow using a locally selected operating threshold and review standard.

Serverless computing has grown rapidly for serving Large Language Model (LLM) inference due to its pay-as-you-go pricing, fine-grained GPU usage, and rapid scaling. However, our analysis reveals that current serverless can effectively serve general LLM but fail with Low-Rank Adaptation (LoRA) inference due to three key limitations: 1) massive parameter redundancy among functions where 99% of weights are unnecessarily duplicated, 2) costly artifact loading latency beyond LLM loading, and 3) magnified resource contention when serving multiple LoRA LLMs. These inefficiencies lead to massive GPU wastage, increased Time-To-First-Token (TTFT), and high monetary costs. We propose ServerlessLoRA, a novel serverless inference system designed for faster and cheaper LoRA LLM serving. ServerlessLoRA enables secure backbone LLM sharing across isolated LoRA functions to reduce redundancy. We design a pre-loading method that pre-loads comprehensive LoRA artifacts to minimize cold-start latency. Furthermore, ServerlessLoRA employs contention aware batching and offloading to mitigate GPU resource conflicts during bursty workloads. Experiment on industrial workloads demonstrates that ServerlessLoRA reduces TTFT by up to 86% and cuts monetary costs by up to 89% compared to state-of-the-art LLM inference solutions.

The rapid adoption of large language models (LLMs) has created significant challenges for efficient inference at scale. Unlike traditional workloads, LLM inference is constrained by both computation and the memory overhead of key-value (KV) caching, which accelerates decoding but quickly exhausts GPU memory. In this paper, we introduce the first queueing-theoretic framework that explicitly incorporates both computation and GPU memory constraints into the analysis of LLM inference. Based on this framework, we derive rigorous stability and instability conditions that determine whether an LLM inference service can sustain incoming demand without unbounded queue growth. This result offers a powerful tool for system deployment, potentially addressing the core challenge of GPU provisioning. By combining an estimated request arrival rate with our derived stable service rate, operators can calculate the necessary cluster size to avoid both costly over-purchasing and performance-violating under-provisioning. We further validate our theoretical predictions through extensive experiments in real GPU production environments. Our results show that the predicted stability conditions are highly accurate, with deviations typically within 10%.

Two-tier LLM serving escalates selected requests from a small model to a large one, but prompt-based routers cannot see whether retrieved evidence supports an answer, and they treat the GPU cost of an escalation as constant. We study TRIAGE, which conditions escalation on a post-generation reliability state and on a load-dependent GPU price, and we report both what held and what did not on one A40 serving Qwen2.5-7B and 32B (AWQ) under VLLM. Escalation cost rose 21.9x from 1 to 10 requests/s but showed no monotone dependence on KV-cache headroom, and a pre-registered claim that the reliability state detects corrupted evidence (AUROC>=0.70) was falsified at 0.571, although any prompt-only router scores exactly 0.500 by construction. Post hoc, a semantic evidence judge raised detection to 0.846 (0.926 on never-served items) and predicted wrong small-model answers at AUROC 0.90-0.96 on RGB and three RouterBench model pairs; the resulting cascade stayed within about 3 pp of GPT-4 at 79.5% lower API cost for GPT-3.5->GPT-4, yet fell below random routing for open-model pairs with 2-3x price ratios. End to end, the 7B->32B pair showed only a +2.88 pp held-out gap, so the allocator rightly declined to escalate and the headline is a null that a small pre-launch gate failed to expose.

Cloud computing is an established technology allowing users to share resources on a large scale, never before seen in IT history. A cloud system connects multiple individual servers in order to process related tasks in several environments at the same time. Clouds are typically more cost-effective than single computers of comparable computing performance. The sheer physical size of the system itself means that thousands of machines may be involved. The focus of this research was to design a strategy to dynamically allocate tasks without overloading Cloud nodes which would result in system stability being maintained at minimum cost. This research has added the following new contributions to the state of knowledge: (i) a novel taxonomy and categorisation of three classes of schedulers, namely OS-level, Cluster and Big Data, which highlight their unique evolution and underline their different objectives; (ii) an abstract model of cloud resources utilisation is specified, including multiple types of resources and consideration of task migration costs; (iii) a virtual machine live migration was experimented with in order to create a formula which estimates the network traffic generated by this process; (iv) a high-fidelity Cloud workload simulator, based on a month-long workload traces from Google's computing cells, was created; (v) two possible approaches to resource management were proposed and examined in the practical part of the manuscript: the centralised metaheuristic load balancer and the decentralised agent-based system. The project involved extensive experiments run on the University of Westminster HPC cluster, and the promising results are presented together with detailed discussions and a conclusion.

Existing large language model (LLM) serving systems fall into two categories: 1) a unified system where prefill phase and decode phase are co-located on the same GPU, sharing the unified computational resource and storage, and 2) a disaggregated system where the two phases are disaggregated to different GPUs. The design of the disaggregated system addresses the latency interference and sophisticated scheduling issues in the unified system but leads to storage challenges including 1) replicated weights for both phases that prevent flexible deployment, 2) KV cache transfer overhead between the two phases, 3) storage imbalance that causes substantial wasted space of the GPU capacity, and 4) suboptimal resource adjustment arising from the difficulties in migrating KV cache. Such storage inefficiency delivers poor serving performance under high request rates. In this paper, we identify that the advantage of the disaggregated system lies in the disaggregated computation, i.e., partitioning the computational resource to enable the asynchronous computation of two phases. Thus, we propose a novel LLM serving system, semi-PD, characterized by disaggregated computation and unified storage. In semi-PD, we introduce a computation resource controller to achieve disaggregated computation at the streaming multi-processor (SM) level, and a unified memory manager to manage the asynchronous memory access from both phases. semi-PD has a low-overhead resource adjustment mechanism between the two phases, and a service-level objective (SLO) aware dynamic partitioning algorithm to optimize the SLO attainment. Compared to state-of-the-art systems, semi-PD maintains lower latency at higher request rates, reducing the average end-to-end latency per request by 1.27-2.58x on DeepSeek series models, and serves 1.55-1.72x more requests adhering to latency constraints on Llama series models.

This paper presents Block, a distributed scheduling framework designed to optimize load balancing and auto-provisioning across instances in large language model serving frameworks by leveraging contextual information from incoming requests. Unlike popular model serving systems that rely on monolithic and heuristic task schedulers, Block operates as a fully distributed, stateless, and predictive scheduling system to achieve low overhead, reliability, and scalability. It leverages the deterministic and predictable characteristics of LLM inferences, such as host configurations, response lengths, and hardware performance, to make scheduling decisions based on accurately predicted metrics. Evaluation on a 12 GPUs cluster shows that Block significantly outperforms heuristic schedulers, boosting serving capacity by up to 16.7\% and reducing P99 tail latency by up to 49.5\%. These performance gains remain consistent across diverse models, workloads and configurations. Code and data are open-sourced.

Existing retrieval-augmented generation (RAG) systems typically use a centralized architecture, causing a high cost of data collection, integration, and management, as well as privacy concerns. There is a great need for a decentralized RAG system that enables foundation models to utilize information directly from data owners who maintain full control over their sources. However, decentralization brings a challenge: the numerous independent data sources vary significantly in reliability, which can diminish retrieval accuracy and response quality. To address this, our decentralized RAG system has a novel reliability scoring mechanism that dynamically evaluates each source based on the quality of responses it contributes to generate and prioritizes high-quality sources during retrieval. To ensure transparency and trust, the scoring process is securely managed through blockchain-based smart contracts, creating verifiable and tamper-proof reliability records without relying on a central authority. We evaluate our decentralized system with two Llama models (3B and 8B) in two simulated environments where six data sources have different levels of reliability. Our system achieves a +10.7\% performance improvement over its centralized counterpart in the real world-like unreliable data environments. Notably, it approaches the upper-bound performance of centralized systems under ideally reliable data environments. The decentralized infrastructure enables secure and trustworthy scoring management, achieving approximately 56\% marginal cost savings through batched update operations. Our code and system are open-sourced at github.com/yining610/Reliable-dRAG.

Large Language Model (LLM) serving has become a critical cloud workload, and realistic traces are essential for motivating and benchmarking serving systems. However, existing LLM serving workload studies remain limited in scale and scope. They often observe short time periods and provide limited visibility into how users interact with models in production. As a result, they do not fully capture how LLM serving workloads evolve over time or how user-model interactions shape production traffic. In this work, we further the understanding of real-world LLM serving workloads through both a global characterization and a longitudinal study of a one-year production trace from CompanyX. Unlike prior studies, our trace captures full production behavior across many models and users, including both popular and long-tail models. We analyze the workload from aggregate, temporal, model-level, and user-level perspectives, revealing workload evolution and user-model structure that are typically hidden behind aggregate views. To support future research, we publicly release the full one-year trace, enabling downstream studies of production behavior without relying on sampled or synthetically generated workloads. The trace is available at https://github.com/HarvardMadSys/chutes_workload.

Multi-turn LLM agents interleave model calls with external tool invocations, shifting serving from stateless request processing to stateful program execution. Serving these workloads requires scheduling, KV-cache management, and routing policies that use program-level context, including turn dependencies, tool-induced gaps, and reusable KV state. Evaluating such policies directly on real systems is costly, since each design point may require dedicated accelerator time across arrival rates, model scales, serving-instance counts, and memory hierarchies. Simulation offers a scalable alternative, but existing LLM serving simulators target stateless request-level workloads and therefore omit the core dynamics of agent serving: multi-turn program execution, cross-turn cache locality, and KV-cache residency during tool gaps. We present AGENTSERVESIM, a hardware-aware simulator for multi-turn LLM agent serving. AGENTSERVESIM evaluates serving policies at program granularity through composable modules: a Program Orchestrator preserves program identity and turn order, a Tool Simulator materializes tool-induced gaps, a Session-Aware Router maintains program-to-instance affinity for cache-aware dispatch, and a KV Residency Model tracks policy-defined KV placement across HBM, host DRAM/CXL, and eviction. Across real serving deployments and hardware configurations, AGENTSERVESIM reproduces real-system behavior within 6% error across key performance metrics while running entirely on commodity CPUs. These results show that AGENTSERVESIM enables controlled, repeatable exploration of agent-serving policies without requiring exhaustive deployment on costly accelerators.

Mixture-of-Experts (MoE) models promise efficient scaling of large language models (LLMs) by activating only a small subset of experts per token, but their parallelized inference pipelines make elastic serving challenging. Existing strategies fall short: horizontal scaling provisions entire replicas of the current configuration, often tens to hundreds of accelerators, leading to coarse granularity, long provisioning delays, and costly overprovisioning. Vertical scaling offers finer adjustments but typically requires instance restarts, incurring downtime. These limitations make current approaches ill-suited for the bursty, short-lived traffic patterns common in cloud deployments. We present ElasticMoE, an elastic scaling framework for MoE LLMs that achieves fine-grained, low-latency, and zero-downtime scaling. ElasticMoE decouples inference execution from memory operations, enabling scaling steps to proceed concurrently with serving. An HBM Management Module (HMM) reuses weights and KV caches via zero-copy remapping, while high-bandwidth peer-to-peer transfers bring newly added accelerators online without interrupting service. A virtual memory based expert redistribution mechanism migrates MoE experts without costly buffer reallocations, reducing peak memory usage during expert parallelism reconfiguration. Our evaluation on Ascend NPUs with three popular MoE LLMs shows that ElasticMoE achieves up to 9x lower scale-up latency, up to 2x better throughput during scaling, and significantly improves SLO attainment compared to baselines. By enabling fine-grained, concurrent scaling with minimal disruption, ElasticMoE advances the practicality of deploying massive MoE LLMs in dynamic cloud environments.

Realistic evaluation of LLM serving systems requires online workloads, dynamic arrivals, queueing, and the serving engine's local scheduling for execution batching, but running such experiments on GPUs is expensive. Existing simulators reduce this cost, but often operate offline or in time-warped mode, re-implement serving-engine schedulers, or require accurate operator/kernel-level latency models. We present LLM-Emu, a serving-native emulator for vLLM that preserves the production HTTP, scheduling, KV-cache, and output-processing paths while replacing only GPU forward execution with profile-sampled latency and synthetic output tokens. Tested on two different GPUs, four model variants, two model families, two attention backends, and both Poisson and bursty ShareGPT workloads, LLM-Emu closely tracks real vLLM serving behavior: TPOT and ITL stay within 4.8% absolute error, E2E latency within 5.3%, and output throughput within 1.9%; TTFT is less stable, with maximum error 10.4%, reflecting its sensitivity to admission and queue state. These results suggest that lightweight, serving-native emulation can support practical online experimentation for LLM-serving systems. LLM-Emu is open sourced at https://github.com/AKafakA/llm-emu.

LLM serving systems process heterogeneous query workloads where different categories exhibit different characteristics. Code queries cluster densely in embedding space while conversational queries distribute sparsely. Content staleness varies from minutes (stock data) to months (code patterns). Query repetition patterns range from power-law (code) to uniform (conversation), producing long tail cache hit rate distributions: high-repetition categories achieve 40-60% hit rates while low-repetition or volatile categories achieve 5-15% hit rates. Vector databases must exclude the long tail because remote search costs (30ms) require 15--20% hit rates to break even, leaving 20-30% of production traffic uncached. Uniform cache policies compound this problem: fixed thresholds cause false positives in dense spaces and miss valid paraphrases in sparse spaces; fixed TTLs waste memory or serve stale data. This paper presents category-aware semantic caching where similarity thresholds, TTLs, and quotas vary by query category. We present a hybrid architecture separating in-memory HNSW search from external document storage, reducing miss cost from 30ms to 2ms. This reduction makes low-hit-rate categories economically viable (break-even at 3-5% versus 15-20%), enabling cache coverage across the entire workload distribution. Adaptive load-based policies extend this framework to respond to downstream model load, dynamically adjusting thresholds and TTLs to reduce traffic to overloaded models by 9-17% in theoretical projections.

Images produced by diffusion models are increasingly popular in digital artwork and visual marketing. However, such generated images might replicate content from existing ones and pose the challenge of content originality. Existing Image Copy Detection (ICD) models, though accurate in detecting hand-crafted replicas, overlook the challenge from diffusion models. This motivates us to introduce ICDiff, the first ICD specialized for diffusion models. To this end, we construct a Diffusion-Replication (D-Rep) dataset and correspondingly propose a novel deep embedding method. D-Rep uses a state-of-the-art diffusion model (Stable Diffusion V1.5) to generate 40, 000 image-replica pairs, which are manually annotated into 6 replication levels ranging from 0 (no replication) to 5 (total replication). Our method, PDF-Embedding, transforms the replication level of each image-replica pair into a probability density function (PDF) as the supervision signal. The intuition is that the probability of neighboring replication levels should be continuous and smooth. Experimental results show that PDF-Embedding surpasses protocol-driven methods and non-PDF choices on the D-Rep test set. Moreover, by utilizing PDF-Embedding, we find that the replication ratios of well-known diffusion models against an open-source gallery range from 10% to 20%.

AI tools are accelerating scientific publication while the systems that review it struggle to keep up, and independent verification of published research has become both harder and more important. As manual replication is slow and expensive, a growing line of work uses coding agents to automate parts of the process. Existing efforts are largely packaged as benchmarks with companion agents that only run inside the benchmark's own pipeline, and no general-purpose replication tool exists. We present VERITAS, a domain-agnostic replication framework built around CLI coding agents. Given a paper, a code repository, or both, VERITAS extracts the paper's claims, runs the methodology while resolving issues as they arise, and judges each claim against the evidence from experiment runs. The pipeline returns an importance-weighted Replication Score, a severity-rated log of every fix applied, and the patched codebase. We evaluate VERITAS on CORE-Bench and ReplicationBench, 65 papers spanning computer science, social science, medicine, and astrophysics. Against two strong Claude Code baselines on the same model and host environment, VERITAS achieves state-of-the-art performance and leads on every metric on both benchmarks.

Large Language Models (LLMs) present a critical trade-off between inference quality and computational cost: larger models offer superior capabilities but incur significant latency, while smaller models are faster but less powerful. Existing serving strategies often employ fixed model scales or static two-stage speculative decoding, failing to dynamically adapt to the varying complexities of user requests or fluctuations in system performance. This paper introduces , a novel framework that reimagines LLM inference as an adaptive routing problem solved through multi-level speculative decoding. dynamically constructs and optimizes inference "paths" (chains of models) based on real-time feedback, addressing the limitations of static approaches. Our contributions are threefold: (1) An adaptive model chain scheduling mechanism that leverages performance profiling (execution times) and predictive similarity metrics (derived from token distribution divergence) to continuously select the optimal sequence of draft and verifier models, minimizing predicted latency per generated token. (2) A multi-level collaborative verification framework where intermediate models within the selected chain can validate speculative tokens, reducing the verification burden on the final, most powerful target model. (3) A synchronized state management system providing efficient, consistent KV cache handling across heterogeneous models in the chain, including precise, low-overhead rollbacks tailored for asynchronous batch processing inherent in multi-level speculation. Preliminary experiments demonstrate the validity of our method.

Oobleck enables resilient distributed training of large DNN models with guaranteed fault tolerance. It takes a planning-execution co-design approach, where it first generates a set of heterogeneous pipeline templates and instantiates at least f+1 logically equivalent pipeline replicas to tolerate any f simultaneous failures. During execution, it relies on already-replicated model states across the replicas to provide fast recovery. Oobleck provably guarantees that some combination of the initially created pipeline templates can be used to cover all available resources after f or fewer simultaneous failures, thereby avoiding resource idling at all times. Evaluation on large DNN models with billions of parameters shows that Oobleck provides consistently high throughput, and it outperforms state-of-the-art fault tolerance solutions like Bamboo and Varuna by up to 29.6x.

Deploying LLMs efficiently requires testing hundreds of serving configurations, but evaluating each one on a GPU cluster takes hours and costs thousands of dollars. Discrete-event simulators are faster and cheaper, but they require re-implementing the serving system's control logic -- a burden that compounds as frameworks evolve. We present Revati, a time-warp emulator that enables performance modeling by directly executing real serving system code at simulation-like speed. The system intercepts CUDA API calls to virtualize device management, allowing serving frameworks to run without physical GPUs. Instead of executing GPU kernels, it performs time jumps -- fast-forwarding virtual time by predicted kernel durations. We propose a coordination protocol that synchronizes these jumps across distributed processes while preserving causality. On vLLM and SGLang, Revati achieves less than 5% prediction error across multiple models and parallelism configurations, while running 5-17x faster than real GPU execution.

Modern LLM serving is no longer homogeneous or monolithic. Production systems now combine disaggregated execution, complex parallelism, runtime optimizations, and stateful workloads such as reasoning, agents, and RL rollouts. Simulation is attractive for exploring this growing design space, yet existing simulators lack the architectural completeness and decision-grade fidelity it demands. Their monolithic-replica abstractions are ill-suited to disaggregated serving, while average-case analytical proxies can distort SLA predictions and even reverse optimization conclusions. We present Frontier, a discrete-event simulator for modern LLM inference serving. Frontier features a disaggregated abstraction. It captures the structure and dynamics of modern serving systems by modeling co-location, Prefill-Decode Disaggregation (PDD), and Attention-FFN Disaggregation (AFD) with role-specific cluster workers, incorporating key runtime optimizations (e.g., CUDA Graphs, speculative decoding) within the scheduler-batch-engine loop, and supporting stateful requests for emerging workloads. It further provides accurate and generalizable predictions of computation, communication, and memory costs across diverse serving scenarios with complex workload compositions. On 16-H800 GPU testbed, Frontier achieves an average throughput error below 4%. Compared with state-of-the-art simulators, it reduces end-to-end latency error from 44.9% to 6.4% under co-location and from 51.7% to 2.6% under disaggregation. It scales to over 1K GPUs on commodity CPUs and enables new use cases such as SLA-dependent Pareto frontier exploration, heterogeneous disaggregated allocation, agentic reasoning scheduling validation, and RL post-training reconfiguration. We release Frontier at https://github.com/NetX-lab/Frontier.

Optimizing embedding systems for production recommendation pipelines demands systematic exploration that consumes disproportionate engineering effort at scale. We apply Andrej Karpathy's AutoResearch paradigm -- a large language model that iteratively edits a training script and retains modifications that improve a held-out scalar metric -- to automate this exploration. We report on twelve weeks of running this paradigm at production scale, where iterations consume hours of multi-GPU compute, evaluation involves competing criteria, and campaigns span weeks across many training jobs. Across two independently developed representation-learning systems for a book recommendation pipeline, we ran 220+ experiments and observed five recurring failure modes absent from the original setting: infrastructure fragility, agent memory decay, search-direction stagnation, iteration-cost asymmetry, and metric fixation. We contribute a three-principle scaffolding design -- prevent, persist, redirect -- that maps each failure mode to a structural remedy and whose instantiation scales with iteration cost. The framework produced a 1.82x Recall@6 lift and a 2.1x coherence lift over hand-tuned baselines, and the agent autonomously designed a text-only fallback that expanded catalog coverage by 5.8x. The two systems span nearly three orders of magnitude in per-iteration cost yet exhibit the same failure modes, suggesting these are structural properties of production-scale autonomous research rather than artifacts of either application.

Serverless computing provides automatic resource management and pay-per-use execution, but effective autoscaling remains challenging because of dynamic workloads, cold-start latency, and dependencies among functions. We present a dependency-aware autoscaling framework that integrates graph-based bottleneck identification, short-term workload forecasting, multi-model consensus, and cost-aware scaling control. Serverless applications are represented as directed dependency graphs, and structurally important functions are identified using weighted degree centrality. Resource demand is predicted using lightweight MLP, LSTM, and CNN models. Their outputs are combined through a performance-weighted probabilistic ensemble inspired by Bayesian model averaging. The controller further incorporates cold-start awareness and cost comparison to select among scale-up, scale-down, and hold actions. Experiments using real workload traces show that supervised forecasting substantially outperforms unsupervised clustering for autoscaling decision generation. The proposed ensemble achieves 99.88 percent prediction accuracy and reduces prediction error compared with representative hybrid forecasting methods. Evaluations across multiple cloud pricing models also demonstrate consistent infrastructure cost reductions while maintaining performance targets. The results show that combining dependency analysis, multi-expert forecasting, and cost-aware control provides a robust and practical solution for serverless autoscaling.

Small open-weight language models are assumed to fail at agentic database work because they lack the reasoning capacity for it. We test that against a production system. Over eleven days we drove the agent mode of an open-source SQL client with 39 open-weight models served locally and one hosted control, across six task surfaces: 8,199 runs, 110,711 ledger events, 14,008 refused tool calls. Of the 2,100 model-attributed agent-mode losses, 1,590, or 75.7%, came from runs that had invoked at least one tool. That majority is what survives resampling models rather than runs: it holds in 99.7% of clustered resamples and in 15 of the 22 models with at least twenty losses. Within it, transport, a run that used the tools and never got a deliverable through, is the largest class at 36.2% and capability, a run that invoked no tool at all, the smallest at 17.3%; we report that ordering as a property of this corpus rather than a general finding, since clustered by model it holds in only 74.5% of resamples. Transport failures decompose into a few mechanical argument shapes. Production ledgers record refusal codes and never the model's arguments, so these were invisible for ten days; capturing them exposed five server defects, one of which demanded a field on one tool, forbade it on the sibling that composed it, then failed the run for its absence. Five server changes, touching no model, prompt or sampling setting, moved six models by 6 to 21 cells out of 30. We also report a confound we believe affects published local-model benchmarks, ours included: with no context cap, one 7.1 GB model was admitted at its full 262,144-token window and held 51 GB on a 64 GB machine, producing runs indistinguishable in any ordinary log from a model timing out. The corpus, the scorer and a verifier that regenerates every figure are released.

Concurrent LLM agents sharing mutable natural-language state produce Structural Race Conditions (SRCs): write-write and cross-shard stale-read conflicts that silently corrupt agent output. Existing multi-agent frameworks (LangGraph, CrewAI, AutoGen) provide no write-ownership semantics over shared state. We present S-Bus, an HTTP middleware whose central mechanism is a server-side DeliveryLog: a per-agent log of HTTP GET operations that automatically reconstructs each agent's read set at commit time without agent SDK changes under HTTP/1.1. The consistency property the DeliveryLog provides -- Observable-Read Isolation (ORI), a partial causal consistency over the HTTP-observable projection of the read set -- prevents structural race conditions when agents collaborate via shared shards. Three contributions: (C1) The DeliveryLog mechanism for automatic HTTP-traffic-based read-set reconstruction, with three-tier mechanised evidence: ReadSetSoundness and ORICommitSafety machine-checked in TLAPS (modulo one retained typing axiom); exhaustive TLC at N=3 (20,763,484 distinct states, zero violations); Dafny discharges 9 inductive soundness lemmas. (C2) Empirical structural-conflict prevention parity against PostgreSQL 17 SERIALIZABLE and Redis 7 WATCH/MULTI on shared-shard contention sweeps with 427,308 active HTTP-409 conflicts: zero Type-I corruptions across all three backends. (C3) ORI's operating envelope is topology-conditional: semantically neutral in dedicated-shard workloads; harmful in single-shard collaborative writing because preservation propagates concurrent contradictions. Source code: https://github.com/sajjadanwar0/sbus

Executable software engineering data is valuable for training SWE agents, but scaling it remains difficult for two reasons: only a small fraction of real repository changes yield verifiable, high-signal task instances, and naively building repository-specific environments quickly becomes the dominant systems cost. We present SWE-Next, an execution-grounded framework for scalable SWE task and trajectory collection. On the data side, SWE-Next mines real merged pull requests, executes candidate base/merged commit pairs, and retains only those that produce strict test improvements without regressions, yielding self-verifying instances. It also applies strict submission gating so that collected trajectories remain evidence-driven rather than speculative. On the systems side, SWE-Next introduces reusable repo-quarter profiles, which reuse the same environment across nearby commits in time while keeping each task run separate and reproducible. Using only 30 hours and 639GB of environment storage, SWE-Next processes 3,971 seed repositories and 102,582 candidate commit pairs mined from real merged PRs to construct a dataset of 2,308 self-verifying instances. Experiments show that SWE-Next improves downstream pass@1 with fewer or comparable training trajectories, indicating that its gains come not from a stronger trajectory generator, but from higher-signal execution-grounded supervision and more efficient data collection.

Existing checkpointing approaches seem ill-suited for distributed training even though hardware limitations make model parallelism, i.e., sharding model state across multiple accelerators, a requirement for model scaling. Consolidating distributed model state into a single checkpoint unacceptably slows down training, and is impractical at extreme scales. Distributed checkpoints, in contrast, are tightly coupled to the model parallelism and hardware configurations of the training run, and thus unusable on different configurations. To address this problem, we propose Universal Checkpointing, a technique that enables efficient checkpoint creation while providing the flexibility of resuming on arbitrary parallelism strategy and hardware configurations. Universal Checkpointing unlocks unprecedented capabilities for large-scale training such as improved resilience to hardware failures through continued training on remaining healthy hardware, and reduced training time through opportunistic exploitation of elastic capacity. The key insight of Universal Checkpointing is the selection of the optimal representation in each phase of the checkpointing life cycle: distributed representation for saving, and consolidated representation for loading. This is achieved using two key mechanisms. First, the universal checkpoint format, which consists of a consolidated representation of each model parameter and metadata for mapping parameter fragments into training ranks of arbitrary model-parallelism configuration. Second, the universal checkpoint language, a simple but powerful specification language for converting distributed checkpoints into the universal checkpoint format. Our evaluation demonstrates the effectiveness and generality of Universal Checkpointing on state-of-the-art model architectures and a wide range of parallelism techniques.

Developing foundation models in medical imaging requires continuous monitoring of downstream performance. Researchers are burdened with tracking numerous experiments, design choices, and their effects on performance, often relying on ad-hoc, manual workflows that are inherently slow and error-prone. We introduce EvalBlocks, a modular, plug-and-play framework for efficient evaluation of foundation models during development. Built on Snakemake, EvalBlocks supports seamless integration of new datasets, foundation models, aggregation methods, and evaluation strategies. All experiments and results are tracked centrally and are reproducible with a single command, while efficient caching and parallel execution enable scalable use on shared compute infrastructure. Demonstrated on five state-of-the-art foundation models and three medical imaging classification tasks, EvalBlocks streamlines model evaluation, enabling researchers to iterate faster and focus on model innovation rather than evaluation logistics. The framework is released as open source software at https://github.com/DIAGNijmegen/eval-blocks.

Retrieval-Augmented Generation (RAG) is often used with Large Language Models (LLMs) to infuse domain knowledge or user-specific information. In RAG, given a user query, a retriever extracts chunks of relevant text from a knowledge base. These chunks are sent to an LLM as part of the input prompt. Typically, any given chunk is repeatedly retrieved across user questions. However, currently, for every question, attention-layers in LLMs fully compute the key values (KVs) repeatedly for the input chunks, as state-of-the-art methods cannot reuse KV-caches when chunks appear at arbitrary locations with arbitrary contexts. Naive reuse leads to output quality degradation. This leads to potentially redundant computations on expensive GPUs and increases latency. In this work, we propose Cache-Craft, a system for managing and reusing precomputed KVs corresponding to the text chunks (we call chunk-caches) in RAG-based systems. We present how to identify chunk-caches that are reusable, how to efficiently perform a small fraction of recomputation to fix the cache to maintain output quality, and how to efficiently store and evict chunk-caches in the hardware for maximizing reuse while masking any overheads. With real production workloads as well as synthetic datasets, we show that Cache-Craft reduces redundant computation by 51% over SOTA prefix-caching and 75% over full recomputation. Additionally, with continuous batching on a real production workload, we get a 1.6X speed up in throughput and a 2X reduction in end-to-end response latency over prefix-caching while maintaining quality, for both the LLaMA-3-8B and LLaMA-3-70B models.

Text-to-image diffusion models (DMs) have achieved remarkable success in image generation. However, concerns about data privacy and intellectual property remain due to their potential to inadvertently memorize and replicate training data. Recent mitigation efforts have focused on identifying and pruning weights responsible for triggering replication, based on the assumption that memorization can be localized. Our research assesses the robustness of these pruning-based approaches. We demonstrate that even after pruning, minor adjustments to text embeddings of input prompts are sufficient to re-trigger data replication, highlighting the fragility of these defenses. Furthermore, we challenge the fundamental assumption of memorization locality, by showing that replication can be triggered from diverse locations within the text embedding space, and follows different paths in the model. Our findings indicate that existing mitigation strategies are insufficient and underscore the need for methods that truly remove memorized content, rather than attempting to suppress its retrieval. As a first step in this direction, we introduce a novel adversarial fine-tuning method that iteratively searches for replication triggers and updates the model to increase robustness. Through our research, we provide fresh insights into the nature of memorization in text-to-image DMs and a foundation for building more trustworthy and compliant generative AI.

Multi-LoRA serving allows many task-specific adapters to share a common base model, but scaling to large adapter fleets introduces substantial storage, loading, and execution overheads. Post-hoc compression can reduce the memory footprint of trained LoRA adapters, yet smaller ranks do not automatically translate into faster serving. For instance, padding, factor access, and generic kernels can obscure the retained low-rank structure, while independently recompressing thousands of modules also incurs significant preparation overhead. We present LoRAForge, a system connecting batched factor construction to rank-specialized execution through an explicit retained-rank representation. It supports two allocation policies: uniform retention ratios (LoRAForge-U) and spectral-energy budget allocation (LoRAForge-S). Both share FlashTSQR factor construction and the same serving backend. FlashTSQR exploits parallelism across modules and row tiles, accelerating factor construction by up to 2.28times over matched library QR. In a fixed-work Qwen control, the shared runtime adds 1.17--1.22times throughput across both compressed policies, with negligible gains on original adapters. Across 85 trained open-weighted Qwen and Llama adapters, the spectral policy's floored quality configuration retains about 7.7% of adapter parameters with fleet-average score drops of at most 1.2 points on a 0--100 scale. The same rank-floored 1/12 spectral exports achieve substantial token-throughput gains on both AMD and NVIDIA GPUs.

Enterprise data lakes often suffer from substantial amounts of duplicate and redundant data, with data volumes ranging from terabytes to petabytes. This leads to both increased storage costs and unnecessarily high maintenance costs for these datasets. In this work, we focus on identifying and reducing redundancy in enterprise data lakes by addressing the problem of 'dataset containment'. To the best of our knowledge, this is one of the first works that addresses table-level containment at a large scale. We propose R2D2: a three-step hierarchical pipeline that efficiently identifies almost all instances of containment by progressively reducing the search space in the data lake. It first builds (i) a schema containment graph, followed by (ii) statistical min-max pruning, and finally, (iii) content level pruning. We further propose minimizing the total storage and access costs by optimally identifying redundant datasets that can be deleted (and reconstructed on demand) while respecting latency constraints. We implement our system on Azure Databricks clusters using Apache Spark for enterprise data stored in ADLS Gen2, and on AWS clusters for open-source data. In contrast to existing modified baselines that are inaccurate or take several days to run, our pipeline can process an enterprise customer data lake at the TB scale in approximately 5 hours with high accuracy. We present theoretical results as well as extensive empirical validation on both enterprise (scale of TBs) and open-source datasets (scale of MBs - GBs), which showcase the effectiveness of our pipeline.

This paper presents a strategy to allocate services on a Cloud system without overloading nodes and maintaining the system stability with minimum cost. We specify an abstract model of cloud resources utilization, including multiple types of resources as well as considerations for the service migration costs. A prototype meta-heuristic load balancer is demonstrated and experimental results are presented and discussed. We also propose a novel genetic algorithm, where population is seeded with the outputs of other meta-heuristic algorithms.

Training compute is increasingly outpacing the availability of high-quality data. This shifts the central challenge from optimal compute allocation to extracting maximum value from limited data. The widely adopted Chinchilla scaling law assumes every training token is unique. This limits its ability to guide pretraining decisions in data-constrained regimes. We model the excess loss under repetition with a simple additive overfitting penalty and find that it accurately describes model behavior. Our scaling law yields qualitatively new compute-optimal allocation advice. Beyond a point, further repetition is counterproductive and compute is better spent on model capacity. We show that following our law's recommended configuration improves performance in data-constrained regimes. Finally, because our one-parameter form isolates overfitting in a single coefficient, it enables direct comparison across training configurations. As a case study, we show that strong weight decay (λ=1.0) reduces this coefficient by approximately 70%, providing a scaling-law explanation for recent findings that optimal weight decay in data-constrained regimes is an order of magnitude larger than standard practice.

Deploying multiple models within shared GPU clusters is a key strategy to improve resource efficiency in large language model (LLM) serving. Existing multi-LLM serving systems improve GPU utilization at the cost of degraded inference performance, particularly time-to-first-token (TTFT). We attribute this degradation to the lack of awareness regarding future workload characteristics. In contrast, recent analyses have shown the strong periodicity and long-term predictability of real-world LLM serving workloads. In this paper, we propose one-for-many GPU prewarming, which proactively loads parameters from multiple models onto GPUs based on workload forecasts. These prewarmed weights enable the system to promptly instantiate serving instances upon encountering request bursts. We design and implement WarmServe, a multi-LLM serving system incorporating three key techniques: (1) a model placement algorithm that optimizes prewarming decisions to minimize cross-model prewarming interference, (2) a KV cache reservation strategy that repurposes idle KV cache space on running GPUs for prewarming new models, and (3) an efficient GPU memory switching mechanism for tensor management. Evaluation on real-world datasets shows that WarmServe reduces tail TTFT by up to 50.8times compared to the state-of-the-art autoscaling-based system, while supporting up to 2.5times higher request throughput than the GPU-sharing system.

Evaluating retrieval-augmented generation (RAG) pipelines requires corpora where ground truth is knowable, temporally structured, and cross-artifact properties that real-world datasets rarely provide cleanly. Existing resources such as the Enron corpus carry legal ambiguity, demographic skew, and no structured ground truth. Purely LLM-generated synthetic data solves the legal problem but introduces a subtler one: the generating model cannot be prevented from hallucinating facts that contradict themselves across documents.We present OrgForge, an open-source multi-agent simulation framework that enforces a strict physics-cognition boundary: a deterministic Python engine maintains a SimEvent ground truth bus; large language models generate only surface prose, constrained by validated proposals. An actor-local clock enforces causal timestamp correctness across all artifact types, eliminating the class of timeline inconsistencies that arise when timestamps are sampled independently per document. We formalize three graph-dynamic subsystems stress propagation via betweenness centrality, temporal edge-weight decay, and Dijkstra escalation routing that govern organizational behavior independently of any LLM. Running a configurable N-day simulation, OrgForge produces interleaved Slack threads, JIRA tickets, Confluence pages, Git pull requests, and emails, all traceable to a shared, immutable event log. We additionally describe a causal chain tracking subsystem that accumulates cross-artifact evidence graphs per incident, a hybrid reciprocal-rank-fusion recurrence detector for identifying repeated failure classes, and an inbound/outbound email engine that routes vendor alerts, customer complaints, and HR correspondence through gated causal chains with probabilistic drop simulation. OrgForge is available under the MIT license.

The rise of LLMs has driven demand for private serverless deployments, characterized by moderate-scale models and infrequent requests. While existing solutions follow exclusive GPU deployment, we take a step back to explore modern platforms and find that: Emerging CPU architectures with built-in accelerators are capable of serving LLMs but remain underutilized, and both CPUs and GPUs can accommodate multiple LLMs simultaneously. We propose LLM-Mesh, a serverless inference scheme for small-to-mid-sized LLMs that enables elastic sharing across heterogeneous hardware. LLM-Mesh tackles three fundamental challenges: (1) precise, fine-grained compute resource allocation at token-level to handle fluctuating computational demands; (2) a coordinated and forward-looking memory scaling mechanism to detect out-of-memory hazards and reduce operational overhead; and (3) a dual approach that reduces resource fragmentation through proactive preemption and reactive bin-packing. Experimental results on 4 32-core CPUs and 4 A100 GPUs show that LLM-Meshimproves service capacity by 44% - 63% through sharing, while further leveraging CPUs boosts this to 91% - 159%.

The growing interest in artificial intelligence has created workloads that require both sequential and random access. At the same time, NVMe-backed storage solutions have emerged, providing caching capability for large columnar datasets in cloud storage. Current columnar storage libraries fall short of effectively utilizing an NVMe device's capabilities, especially when it comes to random access. Historically, this has been assumed an implicit weakness in columnar storage formats, but this has not been sufficiently explored. In this paper, we examine the effectiveness of popular columnar formats such as Apache Arrow, Apache Parquet, and Lance in both random access and full scan tasks against NVMe storage. We argue that effective encoding of a column's structure, such as the repetition and validity information, is the key to unlocking the disk's performance. We show that Parquet, when configured correctly, can achieve over 60x better random access performance than default settings. We also show that this high random access performance requires making minor trade-offs in scan performance and RAM utilization. We then describe the Lance structural encoding scheme, which alternates between two different structural encodings based on data width, and achieves better random access performance without making trade-offs in scan performance or RAM utilization.

Hybrid language models like Jamba mix attention layers with State Space Models (SSMs), creating two memory cache types with opposite profiles: Key-Value (KV) caches grow linearly with sequence length, while SSM states stay fixed per layer. Current inference engines handle this poorly. Unified pools pad SSM states to attention page sizes, wasting up to 7.3x capacity. Static dual pools cannot adapt when prompt distributions shift between requests. We present Asymmetric Virtual Memory Paging (AVMP). The allocator separates the two cache types into physically distinct pools behind a unified virtual address space, and migrates capacity between pools when one runs out. Migration triggers only on allocation failure, keeping behavior deterministic. We evaluate AVMP across 270 synthetic cells plus 60 cells of ShareGPT trace replay on an RTX 3060 12GB. Out-of-Memory events drop 7.6% and request throughput improves 1.83x to 13.3x across synthetic workloads and 2.36x on ShareGPT. All gains hold under paired-bootstrap 95% confidence intervals. A phase-time breakdown reveals two distinct mechanisms: shorter OOM recovery on capacity-pressured workloads, and faster allocation calls on KV-heavy workloads. Implementation is pure Python; Triton integration is future work.

Efficiently serving Large Language Models (LLMs) requires selecting an optimal parallel execution plan, balancing computation, memory, and communication overhead. However, determining the best strategy is challenging due to varying parallelism techniques (data, pipeline, tensor) and workload characteristics (e.g., compute-intensive tasks with long prompts vs. memory-intensive tasks with long generation). We propose APEX, an LLM serving system simulator that efficiently identifies optimal parallel execution plans by considering key factors of LLM serving systems, such as memory usage, batching behavior, etc. APEX performs dynamism-aware simulation to model iteration-level batching, and leverages LLMs' repetitive structure to reduce design space, scaling efficiently to trillion-scale models. APEX abstracts the key components of LLM serving systems, including the model, batching module, quantization formats, and device clusters, enabling the simulator to be general and extensible. Simulating on a CPU, APEX evaluates execution plans for various device clusters, covering diverse LLMs and workloads. APEX finds plans up to 3.37x faster than heuristics, and also plans that reduce energy consumption by up to 45% compared to latency-optimal plans. APEX performs comprehensive evaluations, reporting key system metrics like time per output token and time to first token, which can help service providers meet SLOs. APEX identifies an optimal plan within 15 minutes on a CPU, making it 71x faster and 1234x more cost-effective than cloud-based GPU deployment. APEX can be accessed at https://github.com/microsoft/apex_plus

In large-scale reinforcement learning (RL) systems with decoupled Trainer-Rollout execution, the Trainer must regularly synchronize policy weights to the Rollout side to limit policy staleness. When inter-node bandwidth is abundant, such synchronization is usually only a small fraction of end-to-end cost. As model size grows, however, the communication demand rises rapidly. In bandwidth-constrained or network-variable deployments -- for example, cross-datacenter or cross-cluster settings, heterogeneous resource pools, and online RL -- weight synchronization can become a dominant bottleneck for throughput and tail latency. We observe that, in mainstream large-model RL training, the locations where parameters actually change are highly sparse at the element level (often 99%+ sparsity). Building on this observation, we propose and implement SparseRL-Sync, which replaces full-weight transfers with a lossless sparse update payload (indices and values) that can be exactly reconstructed on the inference side, thereby preserving 100% fidelity. Under a simplified cost model, sparse synchronization reduces the per-update communication volume from S to approximately S/X; with 99% sparsity (X ~ 100), this yields about a 100x reduction in transmitted data. Combined with appropriate bucketing, SparseRL-Sync also reduces launch and control-plane overhead, significantly improving scalability and end-to-end efficiency in bandwidth-limited and highly asynchronous RL settings.

NVIDIA Grace--Blackwell systems make CPU--GPU memory increasingly unified at the programming level, but unified addressability does not make the memory system uniform. A CUDA managed allocation on GB200 may be served from GPU-local HBM, accessed through host/system memory, demand-migrated, prefetched, reused after GPU access, or evicted and reloaded under capacity pressure. The problem is that these states share a programming interface while exercising different hardware paths and runtime mechanisms, making a single ``Unified Memory bandwidth'' number misleading. This paper presents a placement-controlled microbenchmark study of CUDA Unified Memory on NVIDIA GB200, using GH200 as a Grace--Hopper baseline and GB10/DGX Spark as a validation platform. Our methodology separates local HBM access, host/system-memory access, explicit movement, managed first touch, prefetch, reuse, access-pattern sensitivity, oversubscription, and CPU--GPU handoff, then summarizes the observations with a measurement-calibrated model of compute, access, migration, eviction, synchronization, and handoff costs. On GB200, a 1~GB STREAM-style triad reaches 3963~GB/s from local HBM but 304~GB/s through host-mapped CPU/system memory; for a 256~MB streaming kernel, cold CPU-initialized managed memory reaches 14.1~GB/s, while managed prefetch and reuse reach 4606~GB/s and 4504~GB/s. Below the capacity boundary, prefetched and reused managed memory remain above 4.6~TB/s, but at 125\% of CUDA-visible memory they fall to 968~GB/s and 1214~GB/s, showing that Grace--Blackwell unified memory must be understood through placement, access history, locality, reuse, and capacity pressure.

Multi-turn LLM serving accumulates dialogue history whose Key-Value (KV) cache grows with every turn and every user, quickly exceeding the model weights themselves and making memory -- not compute -- the binding constraint on throughput. Non-uniform KV compression, which allocates heterogeneous budgets across attention heads, preserves accuracy far better than uniform schemes, yet remains impractical: modern serving stacks assume identical KV lengths across heads, so heterogeneity traps freed memory as page fragmentation, spends up to 25% of prefill time reclaiming scattered pages, and skews GPU workloads that inflate decode latency by up to 1.7times or burn 15--20% of each decode step on re-planning. We observe that this heterogeneity need not be discovered at runtime: head-wise retention follows a two-level structural regularity -- an input-invariant head ranking with narrowly bounded per-head ratios -- that can be calibrated offline from as few as 50 samples. Building on this insight, we present Tangram, a serving framework that statically resolves what prior systems handle dynamically: Budget Reservation fixes each head's post-compression footprint at scheduling time, eliminating page reclamation; Ragged Paging clusters similar-budget heads into independent page tables, turning fragmentation into reclaimable memory; and Ahead-of-Time Load Balancing precomputes balanced GPU partitions with zero runtime planning. Implemented on vLLM, Tangram serves as a drop-in substrate for existing non-uniform compression methods, matching their accuracy while improving end-to-end throughput by up to 2.6times over the full-KV baseline. Our implementation is publicly available at https://github.com/aiha-lab/TANGRAM.

This paper presents MoE-Infinity, a cost-efficient mixture-of-expert (MoE) serving system that realizes activation-aware expert offloading. MoE-Infinity features sequence-level expert activation tracing, a new approach adept at identifying sparse activations and capturing the temporal locality of MoE inference. By analyzing these traces, MoE-Infinity performs novel activation-aware expert prefetching and caching, substantially reducing the latency overheads usually associated with offloading experts for improved cost performance. Extensive experiments in a cluster show that MoE-Infinity outperforms numerous existing systems and approaches, reducing latency by 4 - 20X and decreasing deployment costs by over 8X for various MoEs. MoE-Infinity's source code is publicly available at https://github.com/TorchMoE/MoE-Infinity

The rapid development of large language models has led to the widespread adoption of Retrieval-Augmented Generation (RAG), which integrates external knowledge to alleviate knowledge bottlenecks and mitigate hallucinations. However, the existing RAG paradigm inevitably suffers from the impact of flawed information introduced during the retrieval phrase, thereby diminishing the reliability and correctness of the generated outcomes. In this paper, we propose Credibility-aware Generation (CAG), a universally applicable framework designed to mitigate the impact of flawed information in RAG. At its core, CAG aims to equip models with the ability to discern and process information based on its credibility. To this end, we propose an innovative data transformation framework that generates data based on credibility, thereby effectively endowing models with the capability of CAG. Furthermore, to accurately evaluate the models' capabilities of CAG, we construct a comprehensive benchmark covering three critical real-world scenarios. Experimental results demonstrate that our model can effectively understand and utilize credibility for generation, significantly outperform other models with retrieval augmentation, and exhibit resilience against the disruption caused by noisy documents, thereby maintaining robust performance. Moreover, our model supports customized credibility, offering a wide range of potential applications.

Mixture-of-Experts (MoE) models challenge serving infrastructures with dynamic, sparse expert utilization, causing instability on conventional systems designed for dense architectures. We propose EaaS, a novel serving system to enable efficient, scalable, and robust MoE deployment. Our system disaggregates MoE modules into independent, stateless services. This design enables fine-grained resource scaling and provides inherent fault tolerance by decoupling compute units. The architecture is powered by a high-performance, CPU-free peer-to-peer communication library that ensures minimal overhead and high throughput. Experiments confirm EaaS's scalability and efficiency, achieving performance comparable to monolithic systems while providing robust fault tolerance and strong scalability. EaaS incurs less than a 2% throughput reduction under simulated hardware failures that would otherwise halt monolithic architectures. It further saves up to 37.5% of computing resources through dynamic fine-grained adaptation to serving traffic, demonstrating strong resilience for large-scale MoE deployment in production.

We present a single-GPU (H100, bf16) evaluation of GPT-OSS-20B (Mixture-of-Experts; 20.9B total, approx. 3.61B active) against dense baselines Qwen3-32B and Yi-34B across multiple dimensions. We measure true time-to-first-token (TTFT), full-decode throughput (TPOT), end-to-end latency percentiles, peak VRAM with past key values (PKV) held, and energy via a consistent nvidia-smi-based sampler. At a 2048-token context with 64-token decode, GPT-OSS-20B delivers higher decode throughput and tokens per Joule than dense baselines Qwen3-32B and Yi-34B, while substantially reducing peak VRAM and energy per 1000 generated tokens; its TTFT is higher due to MoE routing overhead. With only 17.3% of parameters active (3.61B of 20.9B), GPT-OSS-20B provides about 31.8% higher decode throughput and 25.8% lower energy per 1000 generated tokens than Qwen3-32B at 2048/64, while using 31.7% less peak VRAM. Normalized by active parameters, GPT-OSS-20B shows markedly stronger per-active-parameter efficiency (APE), underscoring MoE's deployment advantages. We do not evaluate accuracy; this is a deployment-focused study. We release code and consolidated results to enable replication and extension.

The rapid growth of Large Transformer-based models, specifically Large Language Models (LLMs), now scaling to trillions of parameters, has necessitated training across thousands of GPUs using complex hybrid parallelism strategies (e.g., data, tensor, and pipeline parallelism). Checkpointing this massive, distributed state is critical for a wide range of use cases, such as resilience, suspend-resume, investigating undesirable training trajectories, and explaining model evolution. However, existing checkpointing solutions typically treat model state as opaque binary blobs, ignoring the ``3D heterogeneity'' of the underlying data structures--varying by memory location (GPU vs. Host), number of ``logical'' objects sharded and split across multiple files, data types (tensors vs. Python objects), and their serialization requirements. This results in significant runtime overheads due to blocking device-to-host transfers, data-oblivious serialization, and storage I/O contention. In this paper, we introduce DataStates-LLM, a novel checkpointing architecture that leverages State Providers to decouple state abstraction from data movement. DataStates-LLM exploits the immutability of model parameters during the forward and backward passes to perform ``lazy'', non-blocking asynchronous snapshots. By introducing State Providers, we efficiently coalesce fragmented, heterogeneous shards and overlap the serialization of metadata with bulk tensor I/O. We evaluate DataStates-LLM on models up to 70B parameters on 256 A100-40GB GPUs. Our results demonstrate that DataStates-LLM achieves up to 4times higher checkpointing throughput and reduces end-to-end training time by up to 2.2times compared to state-of-the-art solutions, effectively mitigating the serialization and heterogeneity bottlenecks in extreme-scale LLM training.

Long-horizon LLM agents act on an append-only interaction history that is re-sent to the model at every step, so the context and its cost grow with steps until the sessions exceed the context window. Existing methods manage the context through context requirement prediction, relying on additional model calls, heuristic rules, or trained policies. However, these predictive approaches introduce runtime overhead, invalidate prefix caches, and permanently discard content with no guarantee of recovery. To overcome these limitations, we introduce ReFold: a training-free rendering layer that preserves the underlying interaction history while compressing only the model's rendered context. It removes two kinds of inter-turn redundancy without an auxiliary predictor: content an earlier turn already displayed, replaced by a stub, and turns the agent itself reports finished, folded into a one-line note. Both operators use chunked rendering, rewriting the cached prefix once every few steps rather than at every step. Every removal is strictly reversible, a wrong removal costs one restore from the history rather than permanent content loss. Because it operates at the rendering layer, ReFold is plug-and-play across standard ReAct-style harnesses. Evaluations across five long-horizon benchmarks and two frontier LLMs demonstrate that ReFold reduces token consumption by up to 2.5x and halves the KV-cache memory per session without degrading task success rates. Under capped context budgets, it avoids up to 92% of forced compactions. Under concurrent serving workloads, it reduces request queuing delays by up to 100%, accelerating inference by up to 1.7x, while cutting inference costs by up to 3.4x.

Extensive system scales (i.e. thousands of GPU/TPUs) and prolonged training periods (i.e. months of pretraining) significantly escalate the probability of failures when training large language models (LLMs). Thus, efficient and reliable fault-tolerance methods are in urgent need. Checkpointing is the primary fault-tolerance method to periodically save parameter snapshots from GPU memory to disks via CPU memory. In this paper, we identify the frequency of existing checkpoint-based fault-tolerance being significantly limited by the storage I/O overheads, which results in hefty re-training costs on restarting from the nearest checkpoint. In response to this gap, we introduce an in-memory fault-tolerance framework for large-scale LLM pretraining. The framework boosts the efficiency and reliability of fault tolerance from three aspects: (1) Reduced Data Transfer and I/O: By asynchronously caching parameters, i.e., sharded model parameters, optimizer states, and RNG states, to CPU volatile memory, Our framework significantly reduces communication costs and bypasses checkpoint I/O. (2) Enhanced System Reliability: Our framework enhances parameter protection with a two-layer hierarchy: snapshot management processes (SMPs) safeguard against software failures, together with Erasure Coding (EC) protecting against node failures. This double-layered protection greatly improves the survival probability of the parameters compared to existing checkpointing methods. (3) Improved Snapshotting Frequency: Our framework achieves more frequent snapshotting compared with asynchronous checkpointing optimizations under the same saving time budget, which improves the fault tolerance efficiency. Empirical results demonstrate that Our framework minimizes the overhead of fault tolerance of LLM pretraining by effectively leveraging redundant CPU resources.

Software modernisation through the migration from monolithic architectures to microservices has become increasingly critical, yet identifying effective service boundaries remains a complex and unresolved challenge. Although numerous automated microservice decomposition frameworks have been proposed, their evaluation is often fragmented due to inconsistent benchmark systems, incompatible metrics, and limited reproducibility, thus hindering objective comparison. This work presents a unified comparative evaluation of state-of-the-art microservice decomposition approaches spanning static, dynamic, and hybrid techniques. Using a consistent metric computation pipeline, we assess the decomposition quality across widely used benchmark systems (JPetStore, AcmeAir, DayTrader, and Plants) using Structural Modularity (SM), Interface Number(IFN), Inter-partition Communication (ICP), Non-Extreme Distribution (NED), and related indicators. Our analysis combines results reported in prior studies with experimentally reproduced outputs from available replication packages. Findings indicate that the hierarchical clustering-based methods, particularly HDBScan, produce the most consistently balanced decompositions across benchmarks, achieving strong modularity while minimizing communication and interface overhead.

Offline batch inference, which leverages the flexibility of request batching to achieve higher throughput and lower costs, is becoming more popular for latency-insensitive applications. Meanwhile, recent progress in model capability and modality makes requests more diverse in compute and memory demands, creating unique opportunities for throughput improvement by resource overlapping. However, a request schedule that maximizes resource overlapping can conflict with the schedule that maximizes prefix sharing, a widely-used performance optimization, causing sub-optimal inference throughput. We present BlendServe, a system that maximizes resource utilization of offline batch inference by combining the benefits of resource overlapping and prefix sharing using a resource-aware prefix tree. BlendServe exploits the relaxed latency requirements in offline batch inference to reorder and overlap requests with varied resource demands while ensuring high prefix sharing. We evaluate BlendServe on a variety of synthetic multi-modal workloads and show that it provides up to 1.44times throughput boost compared to widely-used industry standards, vLLM and SGLang.

Data-intensive applications, ranging from large-scale retrieval systems to advanced data pipelines, are increasingly bottlenecked by the processing of highly redundant text corpora. We present Merlin, a local-first, agnostic, high-throughput deduplication and context optimization engine designed to mitigate these inefficiencies. Utilizing a highly optimized, SIMD-friendly open-addressing flat hash set combined with xxHash3-64, Merlin performs rapid, byte-exact deduplication of text passages and data chunks. While broadly applicable to any text-processing workflow, its impact is particularly pronounced in Large Language Model (LLM) ecosystems, such as Retrieval-Augmented Generation (RAG). Our empirical evaluations demonstrate an input reduction ranging from 13.9% in low-redundancy datasets to over 71% in high-redundancy pipelines, maintaining absolute data fidelity. Furthermore, we detail the system's integration architecture via the Model Context Protocol (MCP), enabling secure, zero-network-interception deployment across major IDEs and autonomous agents. This paper outlines the core algorithmic design, performance benchmarks, and the architectural principles required to process data at sustained speeds of up to 8.7 GB/s.

Large language models increasingly serve as persistent knowledge workers, with in-context memory - facts stored in the prompt - as the default strategy. We benchmark in-context memory against Knowledge Objects (KOs), discrete hash-addressed tuples with O(1) retrieval. Within the context window, Claude Sonnet 4.5 achieves 100% exact-match accuracy from 10 to 7,000 facts (97.5% of its 200K window). However, production deployment reveals three failure modes: capacity limits (prompts overflow at 8,000 facts), compaction loss (summarization destroys 60% of facts), and goal drift (cascading compaction erodes 54% of project constraints while the model continues with full confidence). KOs achieve 100% accuracy across all conditions at 252x lower cost. On multi-hop reasoning, KOs reach 78.9% versus 31.6% for in-context. Cross-model replication across four frontier models confirms compaction loss is architectural, not model-specific. We additionally show that embedding retrieval fails on adversarial facts (20% precision at 1) and that neural memory (Titans) stores facts but fails to retrieve them on demand. We introduce density-adaptive retrieval as a switching mechanism and release the benchmark suite.

Two widely adopted techniques for LLM inference serving systems today are hybrid batching and disaggregated serving. A hybrid batch combines prefill and decode tokens of different requests in the same batch to improve resource utilization and throughput at the cost of increased latency per token. In contrast, disaggregated serving decouples compute-bound prefill and bandwidth-bound decode phases to optimize for service level objectives (SLOs) at the cost of resource under-utilization and KV-cache transfer overheads. To address the limitations of these techniques, we propose RAPID-Serve: a technique to concurrently execute prefill and decode on the same GPU(s) to meet latency SLOs while maintaining high throughput and efficient resource utilization. Furthermore, we propose Adaptive Resource Management for runtime compute resource allocation, optionally leveraging CU masking (a fine-grained Compute Unit partitioning feature on AMD InstinctTM GPUs). RAPID-Serve provides up to 4.1x (average 1.7x) unconstrained throughput improvement and 32x and higher (average 4.9x) throughput improvement under SLO constraints, showing it as an effective strategy compared to the state-of-the-art approaches, particularly in resource-constrained environments.

Agentic reinforcement learning (RL) disaggregates training from rollout, so each policy update must reach the rollout clusters before the next batch. Transferring a full 1T checkpoint for such weight synchronization (refit) takes 87.5 min between two AWS regions. Measurements of BF16 training show that about 1% of weights change their stored values per step. Recent systems exploit this sparsity but fall short on placement, exactness, or efficiency: they reimplement placement rules, assemble full tensors, rebuild values arithmetically, or use a cross-cluster collective, and none fully recovers from mid-refit failures. We present NeMo-DCR (Delta-Compressed Refit), which sends only changes yet is bit-exact: receivers obtain the same parameter and buffer bits as a dense refit. For placement, fixed affine mappings project changes from training shards into the checkpoint's canonical coordinates, residual conversion covers the other changes, and the serving runtime's native loader places all changes in receiver storage. For exactness, compressible XOR masks carry affine changes whose projection and loader preserve stored bits, and overwrites carry the others. Receivers apply both in place, retries overwrite partial writes, and a joint commit binds the policy to the baseline for the next delta. For efficiency, object storage or a relay tree streams payloads during delta construction, without a cross-cluster collective. Even at 3% and 5% change rates, NeMo-DCR refits of 30B-1T models are 12-40times faster than a transport-only full-checkpoint reference. A 1T relay-tree refit at 3% takes 150 s instead of 87.5 min, making refits practical for cross-cluster agentic RL at trillion-parameter scale.

Multi-agent applications often execute complex tasks as multi-stage workflows, where each stage is an LLM call whose output becomes part of context for subsequent steps. Existing LLM serving systems largely assume homogeneous clusters with identical model replicas. This design overlooks the potential of heterogeneous deployments, where models of different sizes and capabilities enable finer trade-offs between latency and performance. However, heterogeneity introduces new challenges in scheduling across models with diverse throughput and performance. We present Chimera, a predictive scheduling system for multi-agent workflow serving on heterogeneous LLM clusters that jointly improves end-to-end latency and task performance. Chimera applies semantic routing to estimate per-model confidence scores for each request, predicts the total remaining output length of the workflow, and estimates per-model congestion using in-flight predicted token volumes for load balancing. We evaluate Chimera on representative agentic workflows for code generation and math reasoning using multiple heterogeneous LLM configurations. Across comparable settings, Chimera traces the best latency-performance frontier, reducing end-to-end latency by 1.2--2.4times and improving task performance by 8.0-9.5 percentage points on average over competitive baselines including vLLM.

Large language model (LLM) serving infrastructures are undergoing a shift toward heterogeneity and disaggregation. Modern deployments increasingly integrate diverse accelerators and near-memory processing technologies, introducing significant hardware heterogeneity, while system software increasingly separates computation, memory, and model components across distributed resources to improve scalability and efficiency. As a result, LLM serving performance is no longer determined by hardware or software choices in isolation, but by their runtime interaction through scheduling, data movement, and interconnect behavior. However, understanding these interactions remains challenging, as existing simulators lack the ability to jointly model heterogeneous hardware and disaggregated serving techniques within a unified, runtime-driven framework. This paper presents LLMServingSim 2.0, a unified system-level simulator designed to make runtime-driven hardware-software interactions in heterogeneous and disaggregated LLM serving infrastructures explicit and analyzable. LLMServingSim 2.0 embeds serving decisions and hardware behavior into a single runtime loop, enabling interaction-aware modeling of batching, routing, offloading, memory, and power. The simulator supports extensible integration of emerging accelerators and memory systems through profile-based modeling, while capturing dynamic serving behavior and system-level effects. We validate LLMServingSim 2.0 against real deployments, showing that it reproduces key performance, memory, and power metrics with an average error of 0.95%, while maintaining simulation times of around 10 minutes even for complex configurations. These results demonstrate that LLMServingSim 2.0 provides a practical bridge between hardware innovation and serving-system design, enabling systematic exploration and co-design for next-generation LLM serving infrastructures.

Concerns about reproducibility in artificial intelligence (AI) have emerged, as researchers have reported unsuccessful attempts to directly reproduce published findings in the field. Replicability, the ability to affirm a finding using the same procedures on new data, has not been well studied. In this paper, we examine both reproducibility and replicability of a corpus of 16 papers on table structure recognition (TSR), an AI task aimed at identifying cell locations of tables in digital documents. We attempt to reproduce published results using codes and datasets provided by the original authors. We then examine replicability using a dataset similar to the original as well as a new dataset, GenTSR, consisting of 386 annotated tables extracted from scientific papers. Out of 16 papers studied, we reproduce results consistent with the original in only four. Two of the four papers are identified as replicable using the similar dataset under certain IoU values. No paper is identified as replicable using the new dataset. We offer observations on the causes of irreproducibility and irreplicability. All code and data are available on Codeocean at https://codeocean.com/capsule/6680116/tree.

Large Language Model (LLM) inference is increasingly constrained by GPU memory capacity rather than compute throughput, driven by growing model sizes and the linear growth of the key-value (KV) cache during autoregressive decoding. Existing approaches mitigate memory pressure by offloading model state and KV tensors to host memory, but incur substantial latency due to limited PCIe bandwidth. We present Harvest, an opportunistic GPU cache management framework that exploits high-bandwidth peer-to-peer GPU interconnects to dynamically place model weights and KV cache in unused GPU memory. Harvest treats peer GPU memory as a transient cache tier, preserving correctness while reducing data movement overhead under dynamic memory availability. We demonstrate significant throughput speedup of more than 2 times by using Harvest to accelerate the retrieval of two widely-used inference components: expert layer weights and KV cache entries.

Large Language Models (LLMs) for Generative AI have achieved remarkable progress, evolving into sophisticated and versatile tools widely adopted across various domains and applications. However, the substantial memory overhead caused by their vast number of parameters, combined with the high computational demands of the attention mechanism, poses significant challenges in achieving low latency and high throughput for LLM inference services. Recent advancements, driven by groundbreaking research, have significantly accelerated progress in this field. This paper provides a comprehensive survey of these methods, covering fundamental instance-level approaches, in-depth cluster-level strategies, emerging scenario directions, and other miscellaneous but important areas. At the instance level, we review model placement, request scheduling, decoding length prediction, storage management, and the disaggregation paradigm. At the cluster level, we explore GPU cluster deployment, multi-instance load balancing, and cloud service solutions. For emerging scenarios, we organize the discussion around specific tasks, modules, and auxiliary methods. To ensure a holistic overview, we also highlight several niche yet critical areas. Finally, we outline potential research directions to further advance the field of LLM inference serving.

LAPS identifies and disaggregates requests with different prompt lengths in LLM serving to reduce TTFT latency. While recent systems have decoupled the prefill and decode stages to improve throughput, they still rely on unified scheduling policies that fail to adapt to heterogeneous workload characteristics. We observe that prompt-length variations lead to distinct performance bottlenecks, motivating an adaptive scheduling strategy. LAPS disaggregates multi-turn long-prefill requests from short-prefill ones and introduces a length-aware smart batching mechanism for short-prefill workloads. It adopts a dual-queue design that supports temporal disaggregation on a single prefill instance or spatial disaggregation across multiple instances. For short-prefill batches, a batch waiting window and CUDA Graph-based clustering mitigate interference from heterogeneous computation, reducing batching delay and lowering average latency. In real multi-turn workloads, LAPS reduces prefill latency by over 30\% compared to vanilla SGLang under prefill-decode disaggregation, and further decreases SLO violations by 28\% in multi-instance deployments with vanilla data-parallel configuration. Compared to the SGLang router with load balancing, it further lowers SLO violations by 12\% in multi-GPU settings. Under high concurrency and mixed-request scenarios, LAPS improves request throughput by 35\% serving Qwen2.5-32B model for prefill instance, demonstrating its effectiveness in optimizing heterogeneous LLM serving workloads.

Large Language Model (LLM) serving systems remain fundamentally fragile, where frequent hardware faults in hyperscale clusters trigger disproportionate service outages in the software stack. Current recovery mechanisms are prohibitively slow, often requiring up to 10 minutes to reinitialize resources and reload massive model weights. We introduce KevlarFlow, a fault tolerant serving architecture designed to bridge the gap between hardware unreliability and service availability. KevlarFlow leverages 1) decoupled model parallelism initialization, 2) dynamic traffic rerouting, and 3) background KV cache replication to maintain high throughput during partial failures. Our evaluation demonstrates that KevlarFlow reduces mean-time-to-recovery (MTTR) by 20x and, under failure conditions, improves average latency by 3.1x, 99th percentile (p99) latency by 2.8x, average time-to-first-token (TTFT) by 378.9x, and p99 TTFT by 574.6x with negligible runtime overhead in comparison to state-of-the-art LLM serving systems.

Every public LLM cost calculator we surveyed treats GPU utilization as a fixed input -- entered by the user, baked in as a preset, or silently assumed at 100% -- never measured against the operator's actual load. We show that this assumption is the dominant source of error: on identical H100 hardware, effective cost spans \0.21 to 15.25 per million output tokens, an underutilization penalty of 2.5-24x across low-to-moderate enterprise loads (1-10 rps) and up to 36.3x near idle -- driven by one operator-controlled variable, offered request rate lambda, which sets in-flight concurrency via Little's Law and which no open-source calculator exposes. Because calculators take utilization as a user-supplied input, any utilization-naive estimate understates true cost by exactly 1/U, systematically mispricing self-hosting -- most severely over-selling it for low-traffic workloads. We propose a measurement methodology that parameterizes the relationship as C_eff = f(H, M, Q, lambda, L), validate it with 42 benchmarks across dense, ultra-sparse MoE, and sparse MoE models, and release vllm-cost-meter, an open-source cost meter that attaches to a live vLLM server and reports real \$/M-tokens against the operator's own traffic. We further show that FP8 quantization benefits the MoE architectures we tested roughly 2.2-2.4x more than the dense model (+69 to +74% vs. +31% peak throughput; n=3, broader validation needed), and our data are consistent with active parameter count, not total model size, being a primary predictor of saturation economics. To rule out single-hardware confounding we repeat the core sweep on A100 80GB PCIe (56 runs): the load-driven spread reproduces at 7.0-11.4x, the active-parameters ordering survives at FP8, and the dense-FP8 advantage inverts on silicon without native FP8 tensor cores -- a hardware-conditional caveat the framework already accommodates.

This paper addresses emerging system-level challenges in heterogeneous retrieval-augmented generation (RAG) serving, where complex multi-stage workflows and diverse request patterns complicate efficient execution. We present HedraRAG, a runtime system built on a graph-based abstraction that exposes optimization opportunities across stage-level parallelism, intra-request similarity, and inter-request skewness. These opportunities are realized through dynamic graph transformations, such as node splitting, reordering, edge addition, and dependency rewiring, applied to wavefronts of subgraphs spanning concurrent requests. The resulting execution plans are mapped onto hybrid CPU-GPU pipelines to improve resource utilization and reduce latency. Evaluations across a wide range of RAG workflows demonstrate speedups exceeding 1.5x and reaching up to 5x over existing frameworks, showcasing the effectiveness of coordinated generation and retrieval in serving environments.

Graph Convolutional Networks (GCNs) are increasingly adopted in large-scale graph-based recommender systems. Training GCN requires the minibatch generator traversing graphs and sampling the sparsely located neighboring nodes to obtain their features. Since real-world graphs often exceed the capacity of GPU memory, current GCN training systems keep the feature table in host memory and rely on the CPU to collect sparse features before sending them to the GPUs. This approach, however, puts tremendous pressure on host memory bandwidth and the CPU. This is because the CPU needs to (1) read sparse features from memory, (2) write features into memory as a dense format, and (3) transfer the features from memory to the GPUs. In this work, we propose a novel GPU-oriented data communication approach for GCN training, where GPU threads directly access sparse features in host memory through zero-copy accesses without much CPU help. By removing the CPU gathering stage, our method significantly reduces the consumption of the host resources and data access latency. We further present two important techniques to achieve high host memory access efficiency by the GPU: (1) automatic data access address alignment to maximize PCIe packet efficiency, and (2) asynchronous zero-copy access and kernel execution to fully overlap data transfer with training. We incorporate our method into PyTorch and evaluate its effectiveness using several graphs with sizes up to 111 million nodes and 1.6 billion edges. In a multi-GPU training setup, our method is 65-92% faster than the conventional data transfer method, and can even match the performance of all-in-GPU-memory training for some graphs that fit in GPU memory.

Large language models make weight storage and memory traffic major inference costs, motivating low-precision formats that represent each weight with only a few bits. Such formats use a scale to map floating-point values into a small codebook; NVFP4 improves local range utilization by letting every 16 E2M1 weights share an E4M3 block scale. Choosing that scale is difficult in GPTQ because quantizing one column updates those that follow, so evaluating a block independently can misestimate its final reconstruction error. Large models pose a second challenge: full-precision weights, calibration activations, and second-order state cannot all remain on one accelerator, while assigning complete layers to devices leaves each time-consuming layer solve serial. We introduce Schur Replay, a scale-selection algorithm that reproduces the GPTQ updates caused by each block scale and scores the resulting block error after accounting for compensation from unquantized columns. Separately, our execution infrastructure keeps only the active layer resident, tiers activations across device, host, and disk, retires full-precision layers after export, and distributes independent output rows across tensor-parallel ranks. Together, the algorithm and infrastructure attain 99.35% and 100.84% question-weighted recovery from BF16 across seven benchmarks on Qwen3.5-397B-A17B and Llama-3.3-70B-Instruct. On the 397B model, the infrastructure reduces measured per-layer time by 15.17times over ModelOpt and 23.14times over LLM Compressor, with lower memory used per GPU.

Serving disaggregated large language models (LLMs) over tens of thousands of xPU devices (GPUs or NPUs) with reliable performance faces multiple challenges. 1) Ignoring the diversity (various prefixes and tidal requests), treating all the prompts in a mixed pool is inadequate. To facilitate the similarity per scenario and minimize the inner mismatch on P/D (prefill and decoding) processing, fine-grained organization is required, dynamically adjusting P/D ratios for better performance. 2) Due to inaccurate estimation on workload (queue status or maintained connections), the global scheduler easily incurs unnecessary timeouts in prefill. 3) Block-fixed device-to-device (D2D) KVCache transfer over cluster-level RDMA (remote direct memory access) fails to achieve desired D2D utilization as expected. To overcome previous problems, this paper proposes an end-to-end system P/D-Serve, complying with the paradigm of MLOps (machine learning operations), which models end-to-end (E2E) P/D performance and enables: 1) fine-grained P/D organization, mapping the service with RoCE (RDMA over converged ethernet) as needed, to facilitate similar processing and dynamic adjustments on P/D ratios; 2) on-demand forwarding upon rejections for idle prefill, decoupling the scheduler from regular inaccurate reports and local queues, to avoid timeouts in prefill; and 3) efficient KVCache transfer via optimized D2D access. P/D-Serve is implemented upon Ascend and MindSpore, has been deployed over tens of thousands of NPUs for more than eight months in commercial use, and further achieves 60\%, 42\% and 46\% improvements on E2E throughput, time-to-first-token (TTFT) SLO (service level objective) and D2D transfer time. As the E2E system with optimizations, P/D-Serve achieves 6.7x increase on throughput, compared with aggregated LLMs.

Retrieval-Augmented Generation (RAG) is typically evaluated under the implicit assumption that the underlying knowledge base (KB) is clean, leaving the behaviour of RAG systems under realistic KB degradation poorly characterised. We introduce RAGStress, a controlled evaluation benchmark for stress-testing RAG systems under systematic KB corruption. The benchmark pairs four naturalistic corruption types (factual corruption, numeric typo, relevance poisoning, and contradiction injection) with three severity levels (subtle, moderate, and obvious) over a single-KB, metadata-filtered experimental design built from 57 MMLU subjects and 182,546 documents. Across 52,500 model-question-condition evaluations, RAGStress reveals that clean retrieval can mask robustness differences, semantic-fidelity corruptions are substantially more harmful than signal-utility perturbations, no-retrieval accuracy does not predict corrupted-retrieval robustness, and mixed-KB accuracy should not be treated as worst-case robustness. We document the benchmark's intended use, supported claims, and limitations, and provide an artifact bundle including generation scripts, corruption prompts, metadata schema, and evaluation code. RAGStress is intended as a controlled stress test for RAG robustness under KB corruption, not as a general model leaderboard.

LLM serving relies on prefix caching to improve inference performance. As growing contexts push key-value (KV) cache footprint far beyond GPU HBM and CPU DRAM capacity, KV cache is increasingly offloaded to NVMe SSDs. Unfortunately, restoring KV cache from SSDs suffers from poor I/O performance and incurs significant GPU stalls. This is primarily because the fragmented GPU memory layout results in a massive number of tiny random I/Os, rendering the low-parallelism CPU a severe bottleneck even with GPU Direct Storage (GDS), which still relies on CPU intervention to initiate each I/O and thus remains CPU-centric. This paper presents Tutti, an efficient SSD-backed KV caching solution that eliminates CPU intervention from the critical data and I/O control paths between HBM and SSDs. At the core of Tutti is a GPU-centric KV cache object store, in which the CPU is only responsible for asynchronously loading I/O kernels once per layer to the GPU. Tutti saturates NVMe SSD bandwidth and reduces GPU stalls to near zero through the following designs: (i) we provide a GPU-native object abstraction that enables bulk KV cache transfers and management; (ii) we re-architect the GPU storage stack by introducing GPU io_uring to support asynchronous GPU direct object I/O; and (iii) we propose slack-aware I/O scheduling to avoid GPU resource contention. We have implemented Tutti and integrated it to vLLM. Extensive evaluation shows that compared to the state-of-the-art GDS-enabled, SSD-backed LMCache, Tutti reduces TTFT by 78.3% under strict SLO constraints and improves the achievable request rate by 2x. The serving cost is reduced by 27%. Tutti achieves nearly the same inference performance as DRAM-backed LMCache, while providing almost infinite capacity.

Reproducing machine learning experiments requires reconstructing procedures from information scattered across papers, released code, and supporting artifacts. For coding agents, reproduction success can therefore depend on their ability to fill in missing details, rather than on the released artifacts alone. We ask whether preserving recovered procedures and decisions improves subsequent independent reproduction. We define a standardized artifact design comprising a refined codebase, environment setup, experiment launchers, and a reproduction guide, and introduce ReproBridge, an agent workflow for constructing these refined reproducible artifacts. ReproBridge-Bench provides 1,029 tasks from 387 experiments across 36 papers, anchored by five human-verified reference scores per task. Agents attempt tasks in fresh sessions with raw code or refined artifacts; a separate Auditor checks execution evidence. Across four evaluated models, refined artifacts raise mean reproduction success from 76.4% to 92.6% and reduce downstream API spending by 19.7% on average. The cross-model success-rate spread narrows from 21.6 to 3.9 percentage points, with all four models achieving mean success rates above 90%. Removing any component reduces success in the ablation study, with the largest losses from code and environment removal. These results suggest that preserving execution knowledge helps even models with lower raw-code success reproduce experiments more reliably and at lower API cost.

Corruption studies, the primary tool for evaluating chain-of-thought (CoT) faithfulness, identify which chain positions are "computationally important" by measuring accuracy when steps are replaced with errors. We identify a systematic confound: for chains with explicit terminal answer statements, the dominant format in standard benchmarks, corruption studies detect where the answer text appears, not where computation occurs. A within-dataset format ablation provides the key evidence: on standard GSM8K chains ending with "the answer is X," removing only the answer statement, preserving all reasoning, collapses suffix sensitivity ~19x at 3B (N=300, p=0.022). Conflicting-answer experiments quantify the causal mechanism: at 7B, CC accuracy drops to near-zero (<=0.02) across five architecture families; the followed-wrong rate spans 0.63-1.00 at 3B-7B and attenuates at larger scales (0.300 at Phi-4-14B, ~0.01 at 32B). A within-stable 7B replication (9.3x attenuation, N=76, p=7.8e-3; Qwen3-8B N=299, p=0.004) provides converging evidence, and the pattern replicates on MATH (DeepSeek-R1-7B: 10.9x suffix-survival recovery). On chains without answer suffixes the same protocol identifies the prefix as load-bearing (Delta=-0.77, p<10^-12). Generation-time probes confirm a dissociation: the answer is not early-determined during generation (early commitment <5%), yet at consumption time model outputs systematically follow the explicit answer text. The format-determination effect persists through 14B (8.5x ratio, p=0.001) and converges toward zero at 32B. We propose a three-prerequisite protocol (question-only control, format characterization, all-position sweep) as a minimum standard for corruption-based faithfulness studies.

Large-scale agentic training and evaluation with large language models (LLMs) rely on isolated, stateful execution environments in which models inspect repositories, invoke tools, execute commands, and interact with task-specific services. These workloads create sandboxes in large bursts, span heterogeneous functionality and isolation requirements, retain state across long interactions, and draw from large image corpora with limited reuse. Supporting them therefore requires an elastic execution platform rather than a single sandbox runtime. This report presents DeepSeek Elastic Compute (DSec), a production sandbox platform that exposes FnCall, container, microVM, and full-VM sandbox backends through a unified SDK. DSec coordinates placement and lifecycle management across the cluster, composes environments from independently versioned layers, combines memory sharing, reclamation, and CPU scheduling for high-density execution, and loads image data on demand from Fire-Flyer File System (3FS), a cluster-wide distributed filesystem. DSec is co-designed with the reinforcement learning (RL) framework, decouples stateful rollout execution from preemptible GPU training, coordinates sandbox lifecycle with training to preserve rollout state while reclaiming idle resources, and mitigates agent misbehavior such as reward hacking. A single production-scale unit of DSec spans around 160 nodes, serving about 3 million sandboxes per day; in production, it supports over 380,000 concurrent sandboxes and sustains over 5,000 sandbox creations per second. Our evaluation and deployment experience show that these mechanisms reduce environment setup and image-distribution overhead, improve memory efficiency, and preserve latency-sensitive performance under high-density overcommit.

As the demand for comprehensive evaluations of diverse model capabilities steadily increases, benchmark suites have correspondingly grown significantly in scale. Despite notable advances in redundancy reduction and subset-level performance prediction, a systematic framework that effectively integrates these methods to ensure both prediction accuracy and ranking consistency is still largely elusive. In this paper, we first perform a sample-level analysis of benchmark redundancy and identify several highly similar samples that can be eliminated. Besides, we frame benchmark compression as an optimization problem with the aim of score reconstruction. Building on these, we then propose EssenceBench, a coarse-to-fine framework utilizing an iterative Genetic Algorithm (GA), which takes the advantages of fitness-based subset search and attribution-based sample search. Compared to previous methods, our approach yields superior compression results with lower reconstruction error and markedly higher efficiency. In particular, on the HellaSwag benchmark (10K samples), our method preserves the ranking of all models shifting within 5% using 25x fewer samples, and achieves 95% ranking preservation shifting within 5% using only 200x fewer samples.

Language models are running out of high-quality training data, and even aggressively deduplicated corpora retain some amount of repetition. Earlier controlled studies predated Chinchilla-style scaling laws and could only measure the cost of repetition indirectly. We revisit repetition in the Chinchilla era, using a fitted no-repetition scaling law to report Compute-Equivalent Gain and Compute-Equivalent Loss. We show that under this modernized paradigm, repetition damage is systematic in three ways. First, holding compute allocated to repeated data constant, eval loss peaks at an intermediate repeat count Rep; repeating a moderately sized subset a moderate number of times damages performance more than repeating a large subset a few times or a small subset many times. Second, the location of this peak is well-fit by a power law in model size; this scaling law reveals that the most damaging number of repeated data grows more quickly than compute. Finally, when repeated documents consume 10\% of the FLOPs budget in a controlled exact-document repetition setting, the compute-equivalent loss can be large: on FineWeb-Edu-Dedup, the most damaging repeat count for a Qwen3-style 344M-parameter model at OT=1 matches the loss of a no-repetition run using 67% of the FLOPs. We demonstrate that these phenomena are not language-model-specific, and can be analytically understood in a simple statistical model: a misspecified linear regression with verbatim duplicates reproduces the same qualitative loss peak, quantifying how such peaks can arise from a statistical tradeoff between memorization and generalization. Our findings add precision to the study of duplication in language models, allowing practitioners to quantify the wasted compute incurred by the presence and repeat structure of duplicates in pretraining corpora.

Modern applications increasingly rely on inference serving systems to provide low-latency insights with a diverse set of machine learning models. Existing systems often utilize resource elasticity to scale with demand. However, many applications cannot rely on hardware scaling when deployed at the edge or other resource-constrained environments. In this work, we propose a model selection and scheduling algorithm that implements accuracy scaling to increase efficiency for these more constrained deployments. We show that existing schedulers that make decisions using profiled model accuracy are biased toward the label distribution present in the test dataset. To address this problem, we propose using ML models -- which we call SneakPeek models -- to dynamically adjust estimates of model accuracy, based on the underlying data. Furthermore, we greedily incorporate inference batching into scheduling decisions to improve throughput and avoid the overhead of swapping models in and out of GPU memory. Our approach employs a new notion of request priority, which navigates the trade-off between attaining high accuracy and satisfying deadlines. Using data and models from three real-world applications, we show that our proposed approaches result in higher-utility schedules and higher accuracy inferences in these hardware-constrained environments.

We survey the large language model (LLM) serving area to understand the intricate dynamics between cost-efficiency and accuracy, which is magnified by the growing need for longer contextual understanding when deploying models at a massive scale. Our findings reveal that works in this space optimize along three distinct but conflicting goals: improving serving context length (C), improving serving accuracy (A), and improving serving performance (P). Drawing inspiration from the CAP theorem in databases, we propose a CAP principle for LLM serving, which suggests that any optimization can improve at most two of these three goals simultaneously. Our survey categorizes existing works within this framework. We find the definition and continuity of user-perceived measurement metrics are crucial in determining whether a goal has been met, akin to prior CAP databases in the wild. We recognize the CAP principle for LLM serving as a guiding principle, rather than a formal theorem, to inform designers of the inherent and dynamic trade-offs in serving models. As serving accuracy and performance have been extensively studied, this survey focuses on works that extend serving context length and address the resulting challenges.

High throughput serving of large language models (LLMs) requires batching sufficiently many requests at a time. However, existing systems struggle because the key-value cache (KV cache) memory for each request is huge and grows and shrinks dynamically. When managed inefficiently, this memory can be significantly wasted by fragmentation and redundant duplication, limiting the batch size. To address this problem, we propose PagedAttention, an attention algorithm inspired by the classical virtual memory and paging techniques in operating systems. On top of it, we build vLLM, an LLM serving system that achieves (1) near-zero waste in KV cache memory and (2) flexible sharing of KV cache within and across requests to further reduce memory usage. Our evaluations show that vLLM improves the throughput of popular LLMs by 2-4times with the same level of latency compared to the state-of-the-art systems, such as FasterTransformer and Orca. The improvement is more pronounced with longer sequences, larger models, and more complex decoding algorithms. vLLM's source code is publicly available at https://github.com/vllm-project/vllm