LLM Inference Acceleration

LLM: Large Language Model

Follow this topic Ask about this topic

Momentum

108 papers in the last four weeks, up 120% on the four weeks before. 1.1% of all new papers.

Jul 13Week of Sep 28

Latest papers 683

Jun 1, 2026cs.CL

LayerRoute: Input-Conditioned Adaptive Layer Skipping via LoRA Fine-Tuning for Agentic Language Models

Agentic language model systems alternate between two structurally distinct step types: structured tool calls (short, deterministic, low perplexity) and open-ended planning/reasoning steps (long, complex, high perplexity). Despite this heterogeneity, current inference systems apply identical compute to every step. We introduce LayerRoute, a lightweight adapter that learns to selectively skip transformer blocks on a per-input basis. LayerRoute augments each of the 24 transformer blocks in Qwen2.5-0.5B-Instruct with: (1) a per-layer router (~897 parameters, Linear(896,1)) that outputs a hard binary gate via the straight-through estimator, and (2) LoRA adapters (rank 8, ~1.08M parameters) on the Q/K/V/O attention projections. The backbone weights remain frozen. A single end-to-end training pass on agentic data (Hermes, Glaive, GSM8K, Turing) with a gate regularisation term forces the system to discover which blocks are skippable per input type. After 3,000 steps (6.4 minutes on an A100 40GB), LayerRoute achieves a 12.91% skip differential: tool calls skip 15.25% of FLOPs while planning steps skip only 2.34%, using only 1.10M trainable parameters (0.22% of the 494M backbone). Quality improves over the base model due to LoRA adaptation, with perplexity delta of -1.29 on tool calls and -1.30 on planning.

Prateek Kumar Sikdar

Accenture Bengaluru, India

LLM Inference Acceleration LLM Agents LLM Routing

May 31, 2026cs.CL

LongAttnComp: Cross-Family Context Compression for Long-Context Reasoning

As real-world applications increasingly require processing inputs of 100k+ tokens, the gap between context length and inference efficiency has become a critical bottleneck. Context compression offers a way to reduce prefill costs while preserving task accuracy. However, existing training-free attention-based methods leave substantial gaps in demanding long-context tasks such as code reasoning. We present LongAttnComp, a long-context adaptation of AttnComp that fine-tunes a lightweight cross-attention scoring layer and introduces tokenlevel chunking, a token-budget top-p algorithm, positional reordering, and a formatagnostic query parser. We further design a two-stage fine-tuning recipe for the compressor: Stage 1 builds a general retrieval foundation from NIAH-style data, and Stage 2 extends it with multi-hop and reasoning data for broader long-context task coverage. On InfiniteBench Code-Debug, LongAttnComp matches or exceeds full-context accuracy, substantially outperforms training-free baselines, and transfers across four target models from three families. On LongBench v2, the two-stage recipe largely closes the Stage 1 gap on multi-document reasoning while preserving Code-Debug performance.

Mengmeng Ji, Ravi Shanker Raju, Jonathan Lingjie Li +1

SambaNova Systems, Inc. San Jose, CA, USA

LLM Compression LLM Fine-Tuning LLM Inference Acceleration Long-Context Inference

May 31, 2026cs.LG

Soft-NBCE: Entropy-Weighted Chunk Fusion for Long-Context

The quadratic complexity of self-attention remains a bottleneck for Large Language Models (LLMs) processing ultra-long contexts. The Naive Bayes Cognitive Engine (NBCE) parallelizes long-context inference by chunking documents and routing to the lowest-entropy chunk at each decoding step. This hard-selection strategy causes semantic fragmentation during cross-chunk reasoning, as abrupt routing changes between adjacent tokens disrupt the model's contextual grounding. We present Soft-NBCE, a lightweight extension that replaces discrete chunk selection with soft entropy-weighted chunk fusion. A temperature-scaled Softmax over predictive entropies assigns continuous weights to all chunks, enabling log-space aggregation across chunk-conditioned distributions. To partially compensate for the conditional independence assumption introduced by chunking, we propose Consistency Distillation, a LoRA-based self-distillation that constrains the chunked logit distribution toward a full-context teacher via KL-divergence. On LongBench multi-hop benchmarks, Soft-NBCE with Consistency Distillation improves consistently over NBCE-style baselines (MuSiQue F1: 0.310 vs.\ 0.275 for Vanilla NBCE; HotpotQA F1: 0.479 vs.\ 0.427) while maintaining retrieval accuracy (NIAH-32K: 0.909) at O(L^2/n) peak memory.

Shihao Ji, Mingyu Li, Zihui Song

Chunjiang Intelligence · Beijing Normal University

Long-Context Language Model Inference LLM Inference Acceleration

May 31, 2026cs.CL

Hybrid Verified Decoding: Learning to Allocate Verification in Speculative Decoding

Large Language Model (LLM) generation remains expensive because autoregressive decoding calls the model once for each new token. Speculative decoding reduces this cost by drafting multiple tokens and verifying them with the target model in one step, but its speedup depends on how many drafted tokens are accepted. Parameter-free draft sources can propose long continuations at low cost in structured and agentic workloads, yet a cache match that looks promising at one generation step may have low payoff at the next. We propose Hybrid Verified Decoding, which predicts the accepted length of a cache draft before verification and uses this payoff estimate to choose between cache verification and a model-based drafter. Across three LLMs and sixteen datasets, Hybrid Verified Decoding is especially effective on agentic workflows, where it outperforms EAGLE3 in every setting with a 2.73x average speedup. Our analysis shows how prompt structure creates cache opportunities, how high-payoff cache drafts concentrate in a small part of the draft space, and how payoff-guided selection reduces sequential decoding work, pointing to runtime draft selection as a promising direction for speculative decoding.

Xin Su, Dawid Majchrowski, Fangyuan Yu +5

Thoughtworks · Nvidia

Speculative Decoding LLM Inference Acceleration

May 30, 2026cs.DC

ViBE: Co-Optimizing Workload Skew and Hardware Variability for MoE Serving

In distributed Mixture-of-Experts (MoE) inference, input-dependent token routing interacts with GPU performance variability to create persistent stragglers under synchronized execution, where the slowest GPU determines layer latency. This performance variability is inherent to modern accelerators: manufacturing variation, power limits, and thermal conditions introduce measurable execution-time differences across nominally identical GPUs. The core challenge is that MoE execution-time imbalance arises from the interaction of workload skew and hardware asymmetry. Token routing produces uneven and layer-varying expert loads, while GPU throughput depends on device-specific operating characteristics and workload intensity. Prior work mitigates routing skew but assumes homogeneous hardware, optimizing token balance rather than execution latency. As a result, even balanced token assignments can leave hardware-induced stragglers unaddressed. Thus, we propose Variability-Informed Binning of Experts (ViBE), a hardware-aware expert placement framework that minimizes execution-time imbalance across GPUs. ViBE combines per-GPU performance modeling with expert activation profiling to assign high-load experts to faster devices and low-load experts to slower ones, reducing layer-level stragglers without modifying model semantics or hardware. Because both workload characteristics and effective GPU throughput can shift across serving conditions, ViBE supports lightweight recalibration under workload/performance drift to refresh its routing and performance estimates when needed. Results show that ViBE consistently reduces execution-time imbalance and improves SLO attainment by 14%, while lowering P90 TTFT by up to 45%. We further show that the impact of hardware variability increases at scale, making variability-aware placement important for efficient, high-utilization LLM serving.

Seokjin Go, Marko Scrbak, Ephrem Wu +2

Georgia Institute of Technology Atlanta, GA, USA · Advanced Micro Devices, Inc. Austin, TX, USA · Advanced Micro Devices, Inc. Santa Clara, CA, USA +1

Expert Load Balancing Mixture-of-Experts Inference LLM Inference Acceleration

May 30, 2026cs.CL

ProactiveLLM: Learning Active Interaction for Streaming Large Language Models

Standard Large Language Models (LLMs) follow a read-then-generate paradigm, causing unnecessary latency and computation. Streaming LLMs alleviate this issue by generating while receiving inputs, but still struggle to decide when to interact with the stream. Existing methods either hard-code interaction timing or rely on costly external alignment signals, such as timing labels, reasoning trajectories, or stronger teachers. In this paper, we propose ProactiveLLM, which achieves active interaction by leveraging the model's endogenous states to guide interaction decisions. The model first learns to perceive semantic sufficiency from partial inputs through two complementary training mechanisms: mask-based streaming modeling and synchronized privileged self-distillation (SPSD). The former applies monotonic random masking to the input during training, simulating progressively revealed streaming inputs and enabling the model to learn local semantic dependencies from partial-input views. The latter aligns the partial-context student view with a full-context teacher view generated by the same evolving model, allowing privileged full-context evidence to guide the student's understanding under incomplete observations. Together, these mechanisms induce endogenous sufficiency cues without requiring external teachers or annotations, providing a versatile foundation for the plug-and-play integration of diverse decision heads. Extensive evaluation across text and speech streaming tasks confirms that ProactiveLLM significantly reduces interaction latency while maintaining quality, validating its capacity for dynamic and active interaction. Code is publicly available at https://github.com/EIT-NLP/StreamingLLM/tree/main/ProactiveLLM.

Junlong Tong, Yao Zhang, Anhao Zhao +3

Eastern Institute of Technology, Ningbo · Shanghai Jiao Tong University · Hong Kong Polytechnic University +1

LLM Inference LLM Inference Acceleration

May 30, 2026cs.AI

Threshold-Based Exclusive Batching for LLM Inference

Mixed batching (MB)--interleaving prefill and decode in a single batch--has become the standard scheduling strategy for large language model (LLM) inference due to its efficiency in maximizing compute and memory utilization. However, through controlled experiments, we find that prefill-decode interference inflates MB's per-step marginal cost above that of pure decode. On the high-bandwidth H200 (4.8 TB/s), this occurs only when decode tokens exceed 80% of the batch; however, on the bandwidth-constrained RTX PRO 6000 (1.792 TB/s), this threshold plummets to just 20%. Consequently, the optimal choice between MB and exclusive batching (EB) fundamentally depends on GPU memory bandwidth, model size, and workload composition. We derive a closed-form condition for this EB-MB performance crossover, along with asymptotically optimal phase-switching thresholds and memory-safe batch sizing for EB. Optimized EB achieves up to 41.9% higher throughput on bandwidth-constrained GPUs, while MB retains its advantage on high-bandwidth hardware with larger models. Our hybrid scheduler EB+ applies this condition online to dynamically switch between EB and MB without manual intervention. Under non-stationary traffic with distribution or concurrency shifts, EB+ attains the highest or near-highest throughput in every setting, outperforming MB by up to 36.4%.

Weifang Zhang, Yuzhou Nie, Bowen Pang +2

The Hong Kong Polytechnic University · UC Santa Barbara · Technology and Engineering Center for Space Utilization, Chinese Academy of Sciences +1

GPU Acceleration LLM Inference LLM Inference Scheduling LLM Inference Acceleration

May 29, 2026cs.CL

AdaptR1: Reinforcement Learning Based Adaptive Interleaved Thinking in Multi-hop Question Answering

Large Language Models (LLMs) have achieved remarkable performance in complex reasoning tasks through Chain-of-Thought (CoT) prompting. However, this approach often leads to ``over-thinking,'' where models generate unnecessarily long reasoning traces for simple queries and incur avoidable inference cost. While recent work has explored adaptive reasoning, existing methods typically make a single query-level decision about whether to reason. This overlooks the dynamic nature of multi-step tasks, where the need for explicit reasoning varies across intermediate stages. To address this limitation, we introduce AdaptR1, a Reinforcement Learning (RL) based framework for adaptive interleaved thinking in multi-hop Question Answering (QA). Unlike previous approaches that require Supervised Fine-Tuning (SFT) for cold-start initialization, AdaptR1 uses a fully RL-based strategy with a quality-gated efficiency reward to dynamically allocate reasoning budgets at each step. Under the Graph-R1 setting, AdaptR1 reduces average think tokens by 69.71%, with a 90.35% reduction on HotpotQA, while maintaining performance comparable to or better than standard baselines. Furthermore, our analysis reveals that overthinking in multi-hop reasoning is not uniformly distributed but occurs predominantly during the initial planning stages, highlighting the effectiveness of step-wise adaptive budget allocation.

Yuxin Wang, Jiahao Lu, Qifeng Wu +5

1Computer Science, Fudan University · Institute of Modern Languages and Linguistics, Fudan University · 3Shanghai Innovation Institute +1

Multi-Hop QA RL for Language Model Reasoning LLM Inference Acceleration Large Reasoning Models

May 29, 2026cs.CL

Speculative Pipeline Decoding: Higher-Accruacy and Zero-Bubble Speculation via Pipeline Parallelism

Speculative Decoding (SD) accelerates low-concurrency LLM inference by employing a draft-then-verify paradigm. However, mainstream methods typically rely on multi-token prediction, which introduces escalating prediction difficulty and serial drafting latency. To address these, we propose Speculative Pipeline Decoding (SPD), a groundbreaking framework that unlocks the true potential of pipeline parallelism. By partitioning the target LLM into

n

pipeline stages, SPD allows LLM to process

n

tokens within single sequence in parallel to accelerate decoding. To continuous fill the pipeline in single sequence decoding, a speculation module aggregates intermediate features across different pipeline depths to predict the next token, executing strictly in parallel with the target model's pipeline step, to realize bounded difficulty, higher acceptance rates, and zero latency bubbles. Our experiments demonstrate that SPD achieves significantly higher theoretical and wall-clock speedup compared to mainstream baselines at moderate pipeline depth, though more aggressive settings require further improvement. Our code is available at https://github.com/yuyijiong/speculative_pipeline_decoding

Yijiong Yu, Huazheng Wang, Shuai Yuan +2

aOregon State University · bDeepSolution

Speculative Decoding LLM Inference Acceleration Parallel Decoding Pipeline Parallelism

May 29, 2026cs.LG

CoMem: Context Management with A Decoupled Long-Context Model

Context management enables agentic models to solve long-horizon tasks through iterative summarization of previous interaction histories. However, this process typically incurs substantial decoding overhead for the extra summarization tokens, which significantly affect the end-to-end response latency at deployment. In this paper, we introduce CoMem, a novel framework that decouples memory management from the primary agent workflow, enabling these processes to execute in parallel. We propose a

k

-step-off asynchronous pipeline that overlaps the memory model's summarization with the agent's inference, effectively masking the latency of context processing. To ensure robustness under this asynchronous setting, we introduce a reward-driven training strategy that aligns the memory model to capture sufficient statistics for the agent's decision-making. Theoretical analysis confirms that CoMem offers a superior efficiency-effectiveness trade-off compared to coupled architectures. Our extensive experimental results on SWE-Bench-Verified show that CoMem provides 1.4x latency improvements upon vanilla long-context solutions while preserving most of the performance. Furthermore, we demonstrate that these latency gains scale favorably with increased system throughput, offering a modular path forward for the independent optimization of agent reasoning and memory compression.

Yuwei Zhang, Chengyu Dong, Shuowei Jin +11

Halıcıoğlu Data Science Institute, University of California, San Diego · Amazon

LLM Inference LLM Inference Acceleration LLM Context Management Agent Memory Management

May 29, 2026cs.LG

Reducing the GPU Memory Bottleneck with Lossless Compression for ML -- Extended

Machine learning (ML) training and inference often process data sets far exceeding GPU memory capacity, forcing them to rely on PCIe for on-demand tensor transfers, causing critical transfer bottlenecks. Lossy compression has been proposed to relieve bottlenecks but introduces workload-dependent accuracy loss, making it complex or even prohibitive to use in existing ML deployments. We explore lossless compression as an alternative that avoids this deployment complexity. We identify where lossless compression can be integrated into ML pipelines while minimizing interference with GPU execution. Based on our findings, we introduce Invariant Bit Packing (IBP), a novel lossless compression algorithm designed to minimize data transfer time for ML. IBP identifies and eliminates invariant bits across groups of tensors, improving throughput through GPU-optimized decompression that leverages warp parallelism, low-overhead bit operations, and asynchronous PCIe transfers. We provide easy-to-use APIs, showcasing them by adding IBP support to GNN training, as well as DLRM and LLM inference frameworks. IBP achieves, on average, 74% faster GNN training, 180% faster DLRM embedding lookup, and 24% faster LLM inference.

Aditya K Kamath, Arvind Krishnamurthy, Marco Canini +1

University of Washington Seattle, WA, USA · KAUST Thuwal, Saudi Arabia

Efficient Neural Network Inference GPU Kernel Optimization LLM Inference Acceleration

May 28, 2026cs.CL

Speculative Decoding Across Languages

Speculative decoding has become a crucial component of large language model (LLM) inference, enabling faster generation by drafting multiple tokens and verifying them in parallel. However, small draft models tend to suffer from disproportionately poor multilingual capabilities. Thus, when generating text in a non-English language, speculative decoding is far less effective. We compare three strategies to improve speculative decoding efficiency for eleven languages: finetuning the draft model on task-specific data (translation); finetuning the draft model on unlabeled monolingual corpora; and training simple n-gram draft models on the same monolingual corpora. We evaluate efficiency on translation (from English into the target language) and the held-out task of story generation. We find that while task-specific distillation can significantly improve efficiency, distilled models generalize poorly to a new task. Meanwhile, n-gram draft models, despite lower acceptance rates, consistently provide large speed-ups due to much faster draft generation.

Nirajan Paudel, Michael Ginn, Luc De Nardi +1

University of Colorado

Multilingual Language Models Speculative Decoding LLM Inference Acceleration Language Model Distillation Language Modeling

May 28, 2026cs.CL

Unlocking the Working Memory of Large Language Models for Latent Reasoning

To improve the reasoning capabilities of large language models, test-time compute is typically scaled by generating intermediate tokens before the final answer. However, this couples reasoning to autoregressive generation and thereby conflates internal computation with external communication. In contrast, human cognition can use working memory to hold and manipulate information internally without the need to externalize intermediate thoughts. Drawing on this principle, we introduce Reasoning in Memory (RiM), a latent reasoning method that replaces the autoregressive generation of reasoning steps with memory blocks. These memory blocks are fixed sequences of special tokens that unlock the working-memory capacity of large language models. Since they are fixed rather than generated, they can be processed in a single forward pass, enabling compute-efficient latent reasoning. To operationalize these memory blocks, we employ a two-stage curriculum. First, we ground them by predicting explicit reasoning steps after each memory block. Second, we discard this step-level supervision and iteratively refine the final answer after each memory block. Our experiments on reasoning benchmarks show that, across language models of different families and sizes, RiM matches or exceeds existing latent reasoning methods while avoiding the autoregressive generation of thoughts. These results demonstrate that large language models can be trained to use working memory as an effective mechanism for latent reasoning.

Lukas Aichberger, Sepp Hochreiter

ELLIS Unit Linz and LIT AI Lab, Institute for Machine Learning, Johannes Kepler University Linz, Austria · NXAI GmbH, Linz, Austria

Memory-Augmented Language Models Efficient Language Model Reasoning LLM Inference Acceleration Implicit Reasoning in Language Models LLM Training Latent Reasoning

May 28, 2026cs.AI

Accelerating Constrained Decoding with Token Space Compression

To guarantee that an LLM's outputs conform to a specified structure, context-free grammar (CFG) decoding engines force the selection of next tokens to produce strings that conform to a given CFG. Current CFG-constrained decoding engines are highly optimized, but still suffer from the inherent costs arising from their massive per-step search space---i.e. the entire token vocabulary. This results in intractably high overhead for more complex CFGs, which is precisely the situation where CFG engines are most useful. In this paper, we introduce CFGzip, an offline technique for compressing the token search space, which massively reduces CFG engine overhead. In experiments, we report latency reduction of up to 75x during batched inference, cutting overhead down to ~1.2-2x on the hardest grammars: with CFGzip, constrained decoding is now possible at scale for complex CFGs

Michael Sullivan, Alexander Koller

Department of Language Science and Technology Saarland Informatics Campus Saarland University, Saarbrücken, Germany

Constrained Decoding LLM Inference Acceleration Grammar-Constrained Decoding

May 28, 2026cs.CL

Domino: Decoupling Causal Modeling from Autoregressive Drafting in Speculative Decoding

Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft tokens but incur sequential overhead, while parallel drafters reduce drafting cost but weaken intra-block dependency modeling. In this paper, we propose Domino, a speculative decoding framework that decouples causal dependency modeling from expensive autoregressive draft execution. Domino first uses a parallel draft backbone to produce preliminary draft distributions for the entire block, and then applies a lightweight Domino head to refine them with prefix-dependent causal information. To stabilize teacher-forced causal encoding, we further introduce a base-anchored training curriculum that first strengthens the parallel backbone and then gradually shifts optimization toward the causally corrected final distribution. Experiments on Qwen3 models show that Domino achieves up to

5.49\times

end-to-end speedup under the Transformers backend and up to

5.8\times

throughput speedup under SGLang serving.

Jianuo Huang, Yaojie Zhang, Qituan Zhang +3

1EPIC Lab, Shanghai Jiao Tong University · School of Software Engineering, HUST · 3UESTC +2

Speculative Decoding LLM Inference Acceleration

May 28, 2026cs.CL

Draft-OPD: On-Policy Distillation for Speculative Draft Models

Speculative decoding accelerates large language model inference by pairing a target model with a lightweight draft model whose proposed tokens are verified in parallel. A common way to build draft models, like EAGLE3 or DFlash is supervised fine-tuning (SFT) on target-generated trajectories. However, we observe that SFT quickly plateaus: the draft model's acceptance length on test data stops improving. The reason is an offline-to-inference mismatch: In SFT, the drafter learns from fixed target-generated trajectories, whereas during speculative decoding it is evaluated on blocks proposed under its own policy. This motivates on-policy distillation (OPD), where the target model supervises the drafter on draft-induced states. Yet OPD remains difficult for draft models, as they cannot reliably roll out complete sequences independently, whereas target-assisted generation makes the collected sequences follow the target distribution and thus eliminates the on-policy signal. We therefore propose Draft-OPD, which uses target-assisted rollout for stable continuations and replays drafting from the verification-exposed error positions. This allows the drafter to learn from target feedback on both accepted and rejected proposals, focusing training on the draft-induced errors that limit speculative acceptance. Experiments show that Draft-OPD achieves over

5\times

lossless acceleration for thinking models across diverse tasks, improving over EAGLE-3 and DFlash by 23% and 13%.

Haodi Lei, Yafu Li, Haoran Zhang +8

Shanghai Jiao Tong University · Shanghai AI Laboratory · The Chinese University of Hong Kong +3

Speculative Decoding LLM Inference Acceleration Language Model Distillation On-Policy Distillation

May 27, 2026cs.LG

Parallax: Parameterized Local Linear Attention for Language Modeling

Large Language Models (LLMs) have become the central paradigm in artificial intelligence, yet the core computational primitive of attention has remained structurally unchanged. Local Linear Attention (LLA) is an attention mechanism derived from nonparametric statistics in the test-time regression framework. In contrast to prior research on efficient attention variants, LLA upgrades the local constant estimate in softmax attention to a local linear estimate, yielding provably superior bias-variance tradeoffs for associative memory. However, LLA has not been scaled in LLM pretraining due to computational and numerical stability concerns. We introduce Parallax, a parameterized Local Linear Attention that is scalable for LLMs. Parallax eliminates the numerical solver in LLA and learns an extra query-like projector that probes the KV covariance. We place Parallax within a family of attention mechanisms connected by the bandwidth, the probe construction and the affine structure. We propose a hardware-aware algorithm that increases the arithmetic intensity over FlashAttention, shifting attention into a more compute bound regime. Our prototype decode kernel matches or outperforms FlashAttention 2/3 across diverse batch sizes and context lengths. We pretrain Parallax at 0.6B and 1.7B scales and find consistent perplexity improvements throughout pretraining with gains that transfer to downstream benchmarks. The advantage persists under both parameter-matched and compute-matched controls, demonstrating a Pareto improvement. We perform careful pretraining ablations and identify a novel phenomenon whereby Muon unlocks the capacity of Parallax. To our knowledge, this is the first empirical demonstration of strong architecture-optimizer codesign for attention mechanisms in the architecture research literature.

Yifei Zuo, Dhruv Pai, Zhichen Zeng +3

Northwestern University · Tilde Research · University of Washington

Language Model Pretraining Self-Attention LLM Inference Acceleration Linear Attention

May 27, 2026cs.LG

From Human Guidance to Autonomy: Agent Skill System for End-to-End LLM Deployment on Spatial NPUs

Spatial neural processing units (NPUs) provide an energy-efficient platform for edge LLM inference, but efficiently deploying an LLM end-to-end on such hardware remains labor-intensive. Although AI coding agents have begun to lower this cost, existing studies have largely focused on single-kernel optimization rather than end-to-end LLM deployment on resource-constrained spatial NPUs. We present a two-stage methodology, instantiated on the AMD XDNA 2 NPU, that progresses from human-guided development to agent autonomy. In the first stage, we develop a reference deployment of Llama-3.2-1B through human-guided agent assistance. The resulting implementation achieves a speedup of 2.2x on prefill and 4.0x on decode over the hand-optimized baseline, with the optimization trajectory and its lessons recorded as structured documentation throughout. In the second stage, we distill the documentation into an agent skill system consisting of eight phases, orchestrating the optimization and debugging skill sets, with numerical correctness strictly enforced at each phase. Using our agent skill system, we autonomously deploy eight additional decoder-only LLMs (Llama-3.2-3B, SmolLM2-1.7B, Qwen2.5-{0.5B, 1.5B, 3B}, Qwen3-{0.6B, 1.7B, 4B}) end-to-end on the AMD XDNA 2 NPU using the open-source compiler stack. To our knowledge, these models have not previously been deployed on AMD NPUs via any open-source software stack. Each deployment completes in 0.5-4 hours of agent wall time with almost no human guidance, and passes the numerical-correctness gates, demonstrating functional generalization to previously unencountered LLMs. Three of the eight match or exceed the sustained performance of our Llama-3.2-1B reference deployment, suggesting that the resulting implementations can be competitive without additional model-specific human engineering.

Jiajie Li, Erwei Wang, Zhiru Zhang +1

Cornell University · AMD

AI Accelerator Inference LLM Agent Orchestration LLM Agent Skill Learning LLM Inference Acceleration

May 27, 2026cs.LG

Inference-Native Zeroth-Order Optimization for LLMs

Zeroth-order (ZO) methods train large language models using only forward passes, yet common ZO execution paths perform substantial work beyond what the algorithm itself requires. To remove this extra execution overhead, we present Infer-ZO, which separates the evaluations required by the algorithm from how they are executed, allowing them to reuse existing inference-engine optimizations. After removing inherited execution work, a complete Infer-ZO step on Qwen3-14B adds only 1.16% wall-clock time over matched inference, bringing ZO execution close to its fundamental inference workload. With Infer-ZO, ZO evaluations run at near-inference cost with frozen base weights and can share GPU batches with ordinary inference requests. In co-serving experiments, foreground and background throughput remain within 1-3% of the corresponding background-request baseline. Across 15 Qwen3, Llama, and OPT models, Infer-ZO achieves 2.10x-9.94x end-to-end step speedups over released LoZO. Code is available at https://github.com/playeriv65/zo-vllm.

Zelin Li, Caiwen Ding

University of Minnesota

LLM Serving Zeroth-Order Optimization Efficient Language Model Training LLM Inference Acceleration

May 27, 2026cs.LG

Augmenting Attention with Exponentially Decaying Memory Improves Query-Aware KV Sparsity

Efficient inference is critical for long-context language models, where attention computation and KV-cache access dominate the cost. Recent work RAT+, introduces a recurrence-augmented attention backbone that enables flexible dilated attention at inference time. In this paper, we investigate whether this exponentially decaying memory can also improve existing query-aware sparse inference methods. Using representative methods including Quest, MoBA, and SnapKV, we show that RAT+ consistently improves accuracy over standard attention across sparse budgets on eight needle-in-a-haystack tasks. We validate these gains both on the released checkpoints from the RAT+ paper and on OLMo2-7B, which we continue pretraining with the added memory module for 10B tokens. Finally, we propose two hypotheses explaining why this memory module benefits query-aware sparse inference and design targeted experiments to support them.

Xiuying Wei, Caglar Gulcehre

EPFL, CLAIRE Lausanne, Switzerland

Memory-Augmented Language Models Self-Attention KV Caching Long-Context Language Model Inference LLM Inference Acceleration Sparse Attention

May 27, 2026cs.LG

RW-TTT: Batched Serving for Request-Owned Test-Time Training State

Test-time training (TTT) adapts an LLM during generation by reading and updating request-owned state, such as fast weights, low-rank deltas, or streaming learner state. This breaks batched LLM serving, which assumes shared static weights: serial execution is correct but slow, while naive batching can corrupt request state. We formulate this problem as read-write TTT serving and present RW-TTT , which tags each decode step with its owner, version, and READ/WRITE effect, batches only compatible phases, and commits updates only to the owner. On one GPU with eight fast-weight InPlace-TTT streams, RW-TTT reaches 274.61 aggregate tok/s, 9.31x over sequential serving and 3.44x over per-stream replicas under the same memory budget. It preserves behavior on RULER, a long-context benchmark, and passes owner/version checks.

Jian Yang, Zhizhuo Kou, Yao Tian +4

1The Hong Kong University of Science and Technology · 2The Chinese University of Hong Kong · 3National University of Singapore

LLM Serving Test-Time Optimization LLM Inference Acceleration Test-Time Training

May 27, 2026cs.LG

Locality-Aware Redundancy Pruning for LLM Depth Compression

Large language models are known to contain representational redundancy across network depth, making depth pruning an effective approach for improving inference efficiency. Existing one-shot pruning methods rely on local layer importance or fixed redundancy assumptions across architectures. We propose Locality-Aware Redundancy Pruning (LoRP), a training-free one-shot depth pruning framework guided by representation locality. We show that inter-layer redundancy can be either localized or globally distributed depending on the LLM architecture. To characterize this phenomenon, we introduce Representation Locality Score (RLS), derived from global inter-layer hidden-state similarity. Using a small calibration set, LoRP computes pairwise layer similarity, clusters layers by representational similarity, and allocates pruning according to residual intra-cluster redundancy. Experiments across diverse LLM families show improvements in both perplexity and downstream task accuracy. Official github repository: https://github.com/daniel-eai/LoRP-Locality-Aware-Redundancy-Pruning/

Vincent-Daniel Yun, Youngrae Kim, Woosang Lim +3

University of Southern California, United States · Neural Superintelligence Lab, MODULABS, Republic of Korea · Seoul National University, Republic of Korea +1

LLM Pruning LLM Compression LLM Inference Acceleration

May 26, 2026cs.CL

UNIQUE: Universal Top-k Sparse Attention for Training-free Inference and Sparsity-aware Training

Long-context inference in large language models (LLMs) is bottlenecked by the linear growth of the self-attention key-value (KV) cache. Top-k sparse attention alleviates this by loading only a small fraction of the KV cache, but accurately and cheaply estimating cache importance, for both training-free use and sparsity-aware training, remains challenging. This paper proposes UNIQUE, a universal top-k sparse attention framework that addresses both requirements and stays consistently effective across LLM modalities. UNIQUE operates at the granularity of KV pages and estimates per-page importance with a simple yet accurate score combining the mean of the page's keys as a representative vector with their standard deviation as an offset term. To further close the train-inference gap, this paper introduces a soft-mask sparsity-aware training scheme that uses the top-k score boundary as a per-query threshold and a sigmoid soft mask around it, requiring neither auxiliary losses nor architectural changes. Experiments on text and speech LLMs show that UNIQUE preserves task performance on long-context benchmarks such as LongBench Pro and on long-form speech recognition, while delivering up to 11.4x attention-kernel speedup over FlashInfer dense attention and at least 5.3x end-to-end decoding speedup over a vLLM-based dense model.

Keqi Deng, Shaoshi Ling, Ruchao Fan +1

Microsoft, USA

Self-Attention Long-Context Language Model Inference LLM Inference Acceleration Sparse Attention KV-Cache Compression

May 26, 2026cs.CL

Pair-In, Pair-Out: Latent Multi-Token Prediction for Efficient LLMs

Long chain-of-thought reasoning has made autoregressive decoding the dominant inference cost of modern large language models. Existing methods target either the input side (latent compression) or the output side (speculative decoding and multi-token prediction, MTP), but the two lines of work have been pursued independently. Moreover, output-side methods must incur an expensive verifier pass to validate the unreliable draft tokens predicted by MTP. To address these issues, we propose \textbf{Pair-In, Pair-Out (PIPO)}, which unifies both sides by viewing a latent compressor and an MTP head as mirror-image operations: the compressor folds two input tokens into one latent representation, while the MTP head unfolds one hidden state into one additional output token. To remove the verifier cost without sacrificing reliability, PIPO trains a lightweight confidence head that decides whether draft tokens should be accepted. We observe that On-Policy Distillation (OPD) naturally matches the rejection-sampling criterion of speculative decoding, so the confidence head can be trained alongside OPD with negligible extra cost. Experiments on AIME 2025, GPQA-Diamond, LiveCodeBench v6, and LongBench v2 with Qwen3.5-4B and 9B backbones show that PIPO improves pass@4 over regular decoding by up to

+7.15

points, while delivering up to

2.64\times

first-token-latency and

2.07\times

per-token-latency speedups. Project Page: GitHub.com/RedAI-Infra/PIPO.

Wenhui Tan, Minghao Li, Xiaoqian Ma +5

Gaoling School of Artificial Intelligence, Renmin University of China · AI Platform, Xiaohongshu Inc. · University of Electronic Science and Technology of China +1

Multi-Token Prediction LLM Compression Speculative Decoding LLM Inference Acceleration

May 26, 2026cs.CL

NestedKV: Nested Memory Routing for Long-Context KV Cache Compression

Long-context language models are limited by the memory footprint of the key-value (KV) cache. Existing training-free KV compression methods usually rank tokens by one importance signal -- attention, recency, layer-wise allocation, or key distinctiveness -- which becomes brittle when useful context is globally distinctive, locally episodic, or immediately relevant. We introduce NestedKV, a key-only KV cache compression method inspired by the Continuum Memory System in Nested Learning. NestedKV maintains global, block-level, and sliding-window key anchors, scores tokens by multi-time-scale cosine anomaly, and combines the resulting rankings with a training-free outer learner using head-adaptive mixing and surprise-gated token routing. The score is paired with adaptive per-head budgets and requires no training or LLM modification. Across RULER (4k--32k), LooGLE, LongBench, LongBench-E, InfiniteBench, and MMLU-Pro on Qwen3 and Llama-3.2 models, NestedKV is strongest when the retained cache is small. On Qwen3-4B, it improves over KeyDiff by up to 19.10 points on RULER and 19.29 on LongBench at

r=0.75

; at

r=0.95

, it retains 37.32 on LongBench versus 17.55 for KeyDiff.

Hong Chen, Xiang Liu, Yubo Gao +5

The Hong Kong University of Science and Technology (Guangzhou) · Jimei University

KV Caching Long-Context Language Modeling LLM Inference Acceleration KV-Cache Compression

May 25, 2026cs.CL

Do Language Models Need Sleep? Offline Recurrence for Improved Online Inference

Transformer-based large language models are increasingly used for long-horizon tasks; however, their attention mechanism scales poorly with context length. To handle this, we study a sleep-like consolidation mechanism in which a model periodically converts recent context into persistent fast weights before clearing its key-value cache. During sleep, the model performs

N

offline recurrent passes over the accumulated context and updates the fast weights in its state-space model (SSM) blocks through a learned local rule. During inference, this shifts extra computation to sleep while preserving the latency of wake-time prediction. We test our method on controlled synthetic tasks, including cellular automata and multi-hop graph retrieval, as well as a realistic math reasoning task, on which a regular transformer as well as SSM-attention hybrid models fail. We then show that increasing sleep duration

N

for our models improves performance, with the largest gains on examples that require deeper reasoning.

Sangyun Lee, Sean McLeish, Tom Goldstein +1

Carnegie Mellon University · University of Maryland

Memory-Augmented Language Models Long-Context Language Model Inference Persistent Memory for Language Models Recurrent Transformers LLM Inference Acceleration

May 25, 2026cs.CL

Selective Latent Thinking: Adaptive Compression of LLM Reasoning Chains

Explicit chain-of-thought (CoT) reasoning substantially improves the reasoning ability of large language models (LLMs), but incurs high inference cost due to lengthy autoregressive traces. Existing latent reasoning methods offer a promising alternative, yet they often treat reasoning as uniformly compressible, causing precision-critical intermediate steps to be overly compressed and thereby degrading reasoning accuracy. In this work, we propose Selective Latent Thinking (SLT), a framework that selectively compresses redundant reasoning spans into latent representations while preserving precision-critical spans as explicit CoT within the same reasoning trajectory. Specifically, SLT first uses a lightweight decoder to anticipate a short upcoming reasoning span, and then applies confidence-based gating to determine the longest span that can be reliably compressed. The accepted span is encoded into a compact latent representation to improve reasoning efficiency, while uncertain or precision-critical reasoning remains in explicit CoT form to preserve accuracy. To learn this selective compression policy, SLT adopts a three-stage training strategy that combines span-level latent compression, reliability-aware future reasoning prediction, and trajectory-level reinforcement learning to optimize the trade-off between answer correctness and reasoning cost. Extensive experiments across four mathematical reasoning benchmarks demonstrate that SLT achieves 22.7% higher accuracy than latent reasoning baselines at comparable compression ratios, while reducing reasoning chain length by 58.4% with only 2.8% accuracy degradation compared to explicit CoT,Our code can be found in https://github.com/hunshi34/SLT.

Hui Xie, Jie Liu, Ziyue Qiao +1

Eindhoven University of Technology, Netherlands · School of Computing and Information Technology, Great Bay University, China

Efficient Language Model Reasoning LLM Inference Acceleration CoT Compression LLM Reasoning Latent Reasoning Latent CoT Reasoning

May 25, 2026cs.DC

Neural Router: Semantic Content Matching for Agentic AI

Large language models (LLMs) can serve as the semantic-matching engine of a content-based publish/subscribe broker for agentic AI across the edge-cloud computing continuum, bridging the vocabulary and modality gaps that defeat keyword and embedding filters. Framed as offline multi-label retrieval over three public datasets spanning social-media, legal, and smart-home sensor domains (six LLMs, seven baselines), our central contribution is a two-crossover cost-accuracy characterisation: an analytical context-window crossover below which a CoverAndMerge compression pipeline reduces LLM invocations, and an empirical discrimination-capacity crossover above which matching accuracy collapses independently of context budget, by a model-dependent factor of parameter count and training generation. Two findings carry practical weight: above the discrimination crossover, compression cannot recover accuracy and only frontier-scale models clear large subscription sets; and there backend choice dominates configuration choice, so model selection, not pipeline tuning, is the primary operator lever. We accompany this with three composable algorithms and a per-cluster Quality-of-Experience framework for autonomic LLM-tier selection.

Lauri Lovén, Abhishek Kumar, Alexander Engelhardt +5

Future Computing Group, University of Oulu, Finland · University of Jyväskylä, Finland · Department of Computer Science, University of Helsinki, Finland +1

Semantic Textual Similarity LLM Inference Acceleration Context Compression

May 25, 2026cs.DC

Fine-Tuning and Serving Gemma 4 31B on Google Cloud TPU: A Technical Comparison with GPU Baselines

We present the first end-to-end demonstration of fine-tuning and serving Google's Gemma 4 31B model on TPU hardware, providing an empirical comparison of TPU and GPU platforms for large language model adaptation. Using LoRA on a Google TPU v5p-8 for training and TPU v6e-8 (Trillium) for inference, we document the full set of code-level adaptations required to port a GPU-native training recipe - built on PyTorch, HuggingFace TRL, and FSDP - to the JAX + Tunix/Qwix stack. These adaptations span mesh configuration, LoRA module naming conventions, sharding annotation corrections, gradient checkpoint, data pipeline restructuring, and a custom Orbax-to-safetensor checkpoint merging procedure. For inference, we detail the vLLM-TPU Docker setup necessary to serve Gemma 4 on v6e-8 and characterize the resulting latency and throughput profile. Compared with a similar-costing 2xH100 GPU baseline under identical hyperparameters, TPU training completes 1.61x faster at 2.12x lower cost. For inference, we cover the vLLM-TPU Docker setup required to serve Gemma 4 on v6e-8 and explain the observed latency and throughput characteristics across a QPS sweep spanning 512 to 16k input tokens. Across both workloads we compare performance and cost against a 2xH100 GPU baseline running identical hyperparameters. The TPU completes training 1.61x faster at 2.12x lower cost. For inference, TPU v6e-8 matches GPU at short context (<=2048 tokens) and decisively outperforms at long context: 66% higher throughput and 23.6x faster TTFT at 4096-token inputs (61 ms vs 1,443 ms at QPS=4). Our work removes a critical gap in the open tooling ecosystem and provides practitioners with a recipe for Gemma 4 Dense 31B deployment on the TPU infrastructure.

Jatin Kishnani, Mayank Goel, Amit Singh +2

AI Accelerator Inference LLM Serving LLM Fine-Tuning LLM Inference Acceleration

May 25, 2026cs.CL

IndexMem: Learned KV-Cache Eviction with Latent Memory for Long-Context LLM Inference

Large Language Models (LLMs) are increasingly expected to operate over long contexts, yet standard softmax attention incurs a KV cache that grows linearly with sequence length, quickly becoming the bottleneck for long context inference. A practical remedy is to evict less important KV entries; however, existing eviction policies are largely heuristic and struggle to capture the rich, input-dependent distribution of token importance. In this work, we introduce a learnable indexer that predicts KV importance, enabling more accurate retention of critical tokens. Meanwhile, naively evicting tokens permanently discards their information, leading to irreversible forgetting and degraded retrieval over long ranges. To address this, we propose a lightweight latent memory module that compresses evicted tokens into a compact, online-updated state and provides residual readouts to compensate for the attention contributions lost through KV eviction. Collectively, our method enables accurate long-context inference under a bounded KV budget, delivering consistent improvements on RULER (4K/16K) across Qwen, Mistral, and Llama models (up to 25 points under aggressive eviction), markedly more stable Needle-in-a-Haystack retrieval, and superior LongBench scores and compression curves compared to existing eviction policies.

Xintong Yang, Hao Gu, Binxing Xu +6

The Hong Kong University of Science and Technology · Zhejiang University

Memory-Augmented Language Models Self-Attention KV Caching Long-Context Language Model Inference KV-Cache Eviction LLM Inference Acceleration