ELDR reduces median TPOT by 5.9-13.9% in PD-disaggregated MoE serving via expert signatures from prefill, K-means partitioning, and locality-band routing with KV-co-indexed signature cache.
super hub Mixed citations
Efficient Memory Management for Large Language Model Serving with PagedAttention
Mixed citation behavior. Most common role is background (62%).
113
Pith papers citing it
1
external citations · Pith
Background
62% of classified citations
abstract
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 2-4$\times$ 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
hub tools
citation-role summary
background 10
method 4
dataset 2
citation-polarity summary
claims ledger
- abstract 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
authors
co-cited works
representative citing papers
SGPO extracts strategies from strong-model responses, builds autonomous and guided trajectories, and applies token-level forward-KL distillation with adaptive weighting to outperform SFT and RL baselines by 2.2 points on math benchmarks.
Under GPU-CC, LLM serving losses come from a serialized VM–GPU bridge, not compute; simple scheduling and loader changes recover most of the gap on Blackwell.
Execution-state capsules enable graph-bound full-state checkpointing and sub-millisecond restore for LLMs including KV and recurrent states, yielding 3.9x-27x TTFT speedups in on-device physical-AI serving.
ART optimizes visual pixel inputs to frozen MLLMs to achieve LoRA-competitive accuracy on math and structured tool-use benchmarks without modifying computational graphs.
Alignment defenses adapted from DPO and GRPO mitigate property inference attacks on LLMs while preserving utility.
A GEMM-centric taxonomy and unified benchmark show static depth pruning as the strongest Pareto-optimal baseline for LLM inference acceleration, with the frontier shifting to dynamic depth then static width pruning as quality loss rises.
TTT-RTL performs per-design test-time RL on an LLM policy with EDA-derived PPA rewards and an adaptive KL controller, reducing geometric-mean PPA product by 65.1% on RTLLM v2.0 and ADP by 59.4% on an industrial FPU unit.
MRMMIA is a multi-recall-probe membership inference attack that extracts signals from chat agent memory and outperforms baselines in black-, gray-, and white-box settings.
ARBITER models reasoning trajectory basins in test-time sampling and uses model-internal signals to correct majority-vote failures, recovering part of the oracle gap on math benchmarks.
AVMP separates KV and SSM cache pools behind unified virtual addressing with failure-triggered migration, cutting OOM events 7.6% and raising throughput 1.83-13.3x on synthetic loads and 2.36x on ShareGPT traces.
A multi-agent pipeline iteratively refines topology optimization outputs to match natural language preferences for branched structures, achieving 60% success rate across replicates in cantilever and phone-stand tasks.
CutVerse benchmark evaluates GUI agents on 186 complex media post-production tasks in seven apps and reports 36% success rate for existing models.
GoR extracts citation DAGs using position, frequency, predecessor links and time, then fine-tunes Qwen2.5-7B on 498 seed papers to generate ideas, claiming SOTA over gpt-4o baselines via LLM judges.
Fine-tuned LLMs trained with reinforcement learning using verifiable rewards produce floor plans that satisfy connectivity and numerical constraints, outperforming prior methods with at least 94% relative improvement in compatibility.
NCCLZ decouples quantization and entropy coding across NCCL stack layers to enable overlapped compression, delivering up to 9.65x speedup over plain NCCL on scientific and training workloads.
Power capping is illusory in LLM decode as memory-bound operation leaves power headroom untouched on 700 W GPUs, while SM clock locking saves up to 32% energy and three DVFS classes appear across attention types.
EEP makes wide expert-parallel MoE serving survive single-rank failures with an 11s recovery pause, 8s reintegration pause, and throughput restored to 95% of pre-fault level within 52s while staying within 4.4% of a fixed-membership baseline in steady state.
Apple MPS transformer decoding shows abrupt latency spikes up to 21x in narrow decoding-budget intervals due to KV cache and execution regime shifts, absent on CPU and CUDA.
DUET improves RLVR by allocating tokens across both prompt selection and rollout length, outperforming full-budget baselines even when using only half the tokens.
vOPD stabilizes on-policy distillation gradients by subtracting a closed-form per-token negative reverse KL baseline as a detached control variate, preserving unbiasedness while lowering variance and matching expensive full-vocabulary methods.
CacheFlow cuts TTFT by 10-62% in batched LLM serving via 3D-parallel KV cache restoration and a two-pointer scheduler that overlaps recompute and I/O.
Stealth Pretraining Seeding plants persistent unsafe behaviors in LLMs via diffuse poisoned web content that activates on precise triggers and evades standard evaluation.
Language models learn to evict KV cache entries end-to-end via reinforcement learning from outcome reward alone, achieving 2-3x cache compression while maintaining accuracy on Countdown, AMC, and AIME tasks.
citing papers explorer
-
ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
ELDR reduces median TPOT by 5.9-13.9% in PD-disaggregated MoE serving via expert signatures from prefill, K-means partitioning, and locality-band routing with KV-co-indexed signature cache.
-
Beyond Trajectory Imitation: Strategy-Guided Policy Optimization for LLM Reasoning
SGPO extracts strategies from strong-model responses, builds autonomous and guided trajectories, and applies token-level forward-KL distillation with adaptive weighting to outperform SFT and RL baselines by 2.2 points on math benchmarks.
-
The Serialized Bridge: Understanding and Recovering LLM Serving Performance under Blackwell GPU Confidential Computing
Under GPU-CC, LLM serving losses come from a serialized VM–GPU bridge, not compute; simple scheduling and loader changes recover most of the gap on Blackwell.
-
Execution-State Capsules: Graph-Bound Execution-State Checkpoint and Restore for Low-Latency, Small-Batch, On-Device Physical-AI Serving
Execution-state capsules enable graph-bound full-state checkpointing and sub-millisecond restore for LLMs including KV and recurrent states, yielding 3.9x-27x TTFT speedups in on-device physical-AI serving.
-
Fine-tuning Multi-modal LLMs with ART: Art-based Reinforcement Training
ART optimizes visual pixel inputs to frozen MLLMs to achieve LoRA-competitive accuracy on math and structured tool-use benchmarks without modifying computational graphs.
-
Alignment Defends LLMs from Property Inference Attacks
Alignment defenses adapted from DPO and GRPO mitigate property inference attacks on LLMs while preserving utility.
-
Beyond FLOPs: Benchmarking Real Inference Acceleration of LLM Pruning under a GEMM-Centric Taxonomy
A GEMM-centric taxonomy and unified benchmark show static depth pruning as the strongest Pareto-optimal baseline for LLM inference acceleration, with the frontier shifting to dynamic depth then static width pruning as quality loss rises.
-
Alpha-RTL: Test-Time Training for RTL Hardware Optimization
TTT-RTL performs per-design test-time RL on an LLM policy with EDA-derived PPA rewards and an adaptive KL controller, reducing geometric-mean PPA product by 65.1% on RTLLM v2.0 and ADP by 59.4% on an industrial FPU unit.
-
MRMMIA: Membership Inference Attacks on Memory in Chat Agents
MRMMIA is a multi-recall-probe membership inference attack that extracts signals from chat agent memory and outperforms baselines in black-, gray-, and white-box settings.
-
ARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time Sampling
ARBITER models reasoning trajectory basins in test-time sampling and uses model-internal signals to correct majority-vote failures, recovering part of the oracle gap on math benchmarks.
-
Asymmetric Virtual Memory Paging for Hybrid Mamba-Transformer Inference
AVMP separates KV and SSM cache pools behind unified virtual addressing with failure-triggered migration, cutting OOM events 7.6% and raising throughput 1.83-13.3x on synthetic loads and 2.36x on ShareGPT traces.
-
TO-Agents: A Multi-Agent AI Pipeline for Preference-Guided Topology Optimization
A multi-agent pipeline iteratively refines topology optimization outputs to match natural language preferences for branched structures, achieving 60% success rate across replicates in cantilever and phone-stand tasks.
-
CutVerse: A Compositional GUI Agents Benchmark for Media Post-Production Editing
CutVerse benchmark evaluates GUI agents on 186 complex media post-production tasks in seven apps and reports 36% success rate for existing models.
-
Graphs of Research: Citation Evolution Graphs as Supervision for Research Idea Generation
GoR extracts citation DAGs using position, frequency, predecessor links and time, then fine-tunes Qwen2.5-7B on 498 seed papers to generate ideas, claiming SOTA over gpt-4o baselines via LLM judges.
-
Generative Floor Plan Design with LLMs via Reinforcement Learning with Verifiable Rewards
Fine-tuned LLMs trained with reinforcement learning using verifiable rewards produce floor plans that satisfy connectivity and numerical constraints, outperforming prior methods with at least 94% relative improvement in compatibility.
-
NCCLZ: Compression-Enabled GPU Collectives with Decoupled Quantization and Entropy Coding
NCCLZ decouples quantization and entropy coding across NCCL stack layers to enable overlapped compression, delivering up to 9.65x speedup over plain NCCL on scientific and training workloads.
-
The Illusion of Power Capping in LLM Decode: A Phase-Aware Energy Characterisation Across Attention Architectures
Power capping is illusory in LLM decode as memory-bound operation leaves power headroom untouched on 700 W GPUs, while SM clock locking saves up to 32% energy and three DVFS classes appear across attention types.
-
Surviving Partial Rank Failures in Wide Expert-Parallel MoE Inference
EEP makes wide expert-parallel MoE serving survive single-rank failures with an 11s recovery pause, 8s reintegration pause, and throughput restored to 95% of pre-fault level within 52s while staying within 4.4% of a fixed-membership baseline in steady state.
-
Non-Monotonic Latency in Apple MPS Decoding: KV Cache Interactions and Execution Regimes
Apple MPS transformer decoding shows abrupt latency spikes up to 21x in narrow decoding-budget intervals due to KV cache and execution regime shifts, absent on CPU and CUDA.
-
DUET: Optimize Token-Budget Allocation for Reinforcement Learning with Verifiable Rewards
DUET improves RLVR by allocating tokens across both prompt selection and rollout length, outperforming full-budget baselines even when using only half the tokens.
-
KL for a KL: On-Policy Distillation with Control Variate Baseline
vOPD stabilizes on-policy distillation gradients by subtracting a closed-form per-token negative reverse KL baseline as a detached control variate, preserving unbiasedness while lowering variance and matching expensive full-vocabulary methods.
-
CacheFlow: Efficient LLM Serving with 3D-Parallel KV Cache Restoration
CacheFlow cuts TTFT by 10-62% in batched LLM serving via 3D-parallel KV cache restoration and a two-pointer scheduler that overlaps recompute and I/O.
-
PermaFrost-Attack: Stealth Pretraining Seeding(SPS) for planting Logic Landmines During LLM Training
Stealth Pretraining Seeding plants persistent unsafe behaviors in LLMs via diffuse poisoned web content that activates on precise triggers and evades standard evaluation.
-
Neural Garbage Collection: Learning to Forget while Learning to Reason
Language models learn to evict KV cache entries end-to-end via reinforcement learning from outcome reward alone, achieving 2-3x cache compression while maintaining accuracy on Countdown, AMC, and AIME tasks.
-
Sparse Prefix Caching for Hybrid and Recurrent LLM Serving
Sparse prefix caching via dynamic programming for optimal checkpoint placement under overlap distributions improves the Pareto frontier for recurrent and hybrid LLM serving on shared-prefix data.
-
LoSA: Locality Aware Sparse Attention for Block-Wise Diffusion Language Models
LoSA caches prefix attention for stable tokens in block-wise DLMs and applies sparse attention only to active tokens, preserving near-dense accuracy while achieving 1.54x lower attention density and up to 4.14x speedup.
-
MURPHY: Feedback-Aware GRPO with Retrospective Credit Assignment for Multi-Turn Code Generation
MURPHY improves code generation pass rates by up to 6% through retrospective credit assignment on multi-turn feedback trees using max or mean reward propagation.
-
SnapStream: Efficient Long Sequence Decoding on Dataflow Accelerators
SnapStream deploys sparse KV attention in a production inference system on dataflow accelerators, delivering 4x on-chip memory savings for DeepSeek-671B at 128k context with up to 1832 tokens/sec and minimal accuracy loss on LongBench-v2, AIME24, and LiveCodeBench.
-
Does Mixture-of-Experts Actually Help Inference on Consumer and Edge Hardware? An Empirical Study
On bandwidth-bound edge hardware, MoE inference cost tracks total parameters rather than active ones, so sparse activation fails to deliver the expected throughput or energy gains.
-
What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents
KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.
-
AI Native Games: A Survey and Roadmap
AI-native games require runtime generative AI as a non-substitutable core-loop mechanism; a 53-game corpus clusters in language-forward narrative and epistemic designs.
-
A Systematic Approach to Multi-Agent AI from Advanced Regulatory Control Theory: Safe and Auditable LLM Operator Agents for Process Control
ARC-derived multi-agent LLM framework for safe, auditable process control with operator agents and deterministic orchestrator, evaluated on dairy ventilation.
-
When Calibration Rankings Reverse: Accuracy-Controlled Evaluation for Fair Comparison of LLMs
Global calibration metrics like ECE are confounded by accuracy; the proposed ACE framework with three accuracy-controlled views shows many prior calibration advantages weaken or reverse.
-
Mechanistically Eliciting Latent Behaviors in Language Models
CPE is an unsupervised tensor-decomposition method that finds interpretable LoRAs to surface hidden LLM behaviors, matching supervised methods on some tasks and revealing failure modes like sandbagging and alignment-faking.
-
CHIA: An open-source framework for principled, agentic AI-driven hardware/software co-design research
CHIA is a distributed graph-based framework for composing, deploying, and verifying agentic AI hardware/software co-design flows, demonstrated on RTL, simulation, and compiler tasks.
-
RKSC: Reasoning-Aware KV Cache Sharing and Confident Early Exit for Multi-Step LLM Inference
RKSC delivers 3.008x mean speedup over baseline and 1.66x over vLLM prefix caching for multi-branch LLM reasoning via similarity-based KV sharing and confidence-gated early exit, with 0.37% error rate.
-
SpectrumKV: Per-Token Mixed-Precision KV Cache Transfer for Prefill-Decode Disaggregated LLM Serving
SpectrumKV applies per-token mixed-precision KV cache transfer (FP16/INT8/INT4) with a model-specific probe for INT4 tolerance, achieving better perplexity and retrieval than PDTrim at equivalent budgets on Qwen2.5-7B, Mistral-7B, and Gemma-2-9B.
-
Harnessing Streaming Video in the Wild
Presents Streaming-Train-248K dataset, Streaming Harness system, and Streaming-Eval benchmark to enable VLMs for proactive, memory-equipped streaming video understanding.
-
vla.cpp: A Unified Inference Runtime for Vision-Language-Action Models
vla.cpp is a unified C++ runtime that serves multiple VLA architectures with flow-matching and diffusion patterns, matching SOTA performance on LIBERO while running on low-memory embedded hardware.
-
SANE Schema-aware Natural-language Evaluation of Biological Data
SANE is a new schema-aware benchmark paradigm for text-to-SQL evaluation that demonstrates few-shot LLMs with structured prompting can generate accurate queries on constrained biological data schemas without fine-tuning.
-
AlignAtt4LLM: Fast AlignAtt for Decoder-Only LLMs at IWSLT 2026 Simultaneous Speech Translation Task
AlignAtt4LLM adapts AlignAtt to decoder-only LLMs via prompt layout, head selection, and attention replay, outperforming IWSLT 2026 baselines for En-De and En-It at ~2s and <4s latency.
-
When Attention Collapses: Stage-Aware Visual Token Pruning from Structure to Semantics
STS is a two-stage pruning framework that decouples structural diversity via repulsion sampling from semantic filtering via cross-attention to reduce redundancy in visual tokens for VLMs.
-
Traj-Evolve: A Self-Evolving Multi-Agent System for Patient Trajectory Modeling in Lung Cancer Early Detection
Traj-Evolve combines non-parametric experience retrieval and multi-agent RL with a leave-one-out unification strategy to outperform baselines on lung cancer prediction from up to five years of multimodal EHRs, including in never-smokers.
-
Planner-Centric Reinforcement Learning for Deep Research with Structure-Aware Reward
DecomposeR represents research plans as typed DAGs and uses two-stage planner-then-answerer RL to improve long-form research performance by 5.1-8.0 points over baselines.
-
A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving
The paper introduces a paired testing protocol for batch-conditioned refusal robustness in LLM serving and reports low rates of genuine safety-label flips after adjudication, with a batch-invariant kernel ablation eliminating observed flips.
-
Tensor Memory: Fixed-Size Recurrent State for Long-Horizon Transformers
Tensor Memory augments Transformers with a constant-size 3D voxel grid using differentiable soft writes at predicted locations, local interaction, and gated recurrent dynamics to decouple memory capacity from sequence length.
-
Identifying and Mitigating Systemic Measurement Bias in Production LLM Inference Benchmarks
Standard LLM inference benchmarks introduce systematic bias via GIL-induced queuing in single-process asyncio setups; a multi-process framework and NTPOT metric isolate true serving engine performance at high query rates.
-
When Are Teacher Tokens Reliable? Position-Weighted On-Policy Self-Distillation for Reasoning
Position-Weighted On-Policy Self-Distillation (PW-OPSD) weights later tokens more heavily after a diagnostic shows position predicts teacher reliability better than entropy, yielding +1.0 and +1.1 Avg@12 gains on AIME 2024/2025.
-
Runtime-Certified Bounded-Error Quantized Attention
A tiered KV cache architecture computes per-head per-step error bounds on quantized attention and uses adaptive fallback to guarantee bounded or exact outputs relative to FP16 reference.
-
The Constraint Tax: Measuring Validity-Correctness Tradeoffs in Structured Outputs for Small Language Models
Enforcing hard schemas on sub-3B models raises schema validity to 100% but drops answer accuracy from 19.7% to 11.0% and executable accuracy from 91.5% to 48.0% on tool-call tasks.