SpecAttn: Speculating Sparse Attention

• URL: http://arxiv.org/abs/2510.27641v1
• Date: Fri, 31 Oct 2025 17:12:34 GMT
• Title: SpecAttn: Speculating Sparse Attention
• Authors: Harsh Shah,
• Abstract summary: We introduce SpecAttn, a novel training-free approach that seamlessly integrates with speculative decoding techniques.<n>Our key insight is to exploit the attention weights already computed by the draft model during speculative decoding to identify important tokens for the target model.<n>SpecAttn achieves over 75% reduction in key-value cache accesses with a mere 15.29% increase in perplexity on the PG-19 dataset.
• Score: 1.6921396880325779
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) face significant computational bottlenecks during inference due to the quadratic complexity of self-attention mechanisms, particularly as context lengths increase. We introduce SpecAttn, a novel training-free approach that seamlessly integrates with existing speculative decoding techniques to enable efficient sparse attention in pre-trained transformers. Our key insight is to exploit the attention weights already computed by the draft model during speculative decoding to identify important tokens for the target model, eliminating redundant computation while maintaining output quality. SpecAttn employs three core techniques: KL divergence-based layer alignment between draft and target models, a GPU-optimized sorting-free algorithm for top-p token selection from draft attention patterns, and dynamic key-value cache pruning guided by these predictions. By leveraging the computational work already performed in standard speculative decoding pipelines, SpecAttn achieves over 75% reduction in key-value cache accesses with a mere 15.29% increase in perplexity on the PG-19 dataset, significantly outperforming existing sparse attention methods. Our approach demonstrates that speculative execution can be enhanced to provide approximate verification without significant performance degradation.

Related papers

• POP: Prefill-Only Pruning for Efficient Large Model Inference [5.743318651374061]
  Large Language Models (LLMs) and Vision-Language Models (VLMs) have demonstrated remarkable capabilities.<n>Existing structured pruning methods, while hardware-efficient, often suffer from significant accuracy degradation.<n>We argue that this failure stems from a stage-agnostic pruning approach that overlooks the asymmetric roles between the prefill and decode stages.
  arXiv  Detail & Related papers  (2026-02-03T09:22:26Z)
• Accelerate Speculative Decoding with Sparse Computation in Verification [49.74839681322316]
  Speculative decoding accelerates autoregressive language model inference by verifying multiple draft tokens in parallel.<n>Existing sparsification methods are designed primarily for standard token-by-token autoregressive decoding.<n>We propose a sparse verification framework that jointly sparsifies attention, FFN, and MoE components during the verification stage to reduce the dominant computation cost.
  arXiv  Detail & Related papers  (2025-12-26T07:53:41Z)
• Training-free Context-adaptive Attention for Efficient Long Context Modeling [57.703159205740185]
  Training-free Context-adaptive Attention (TCA-Attention) is a training-free sparse attention mechanism that selectively attends to only the informative tokens for efficient long-context inference.<n>TCA-Attention achieves a 2.8$times$ speedup and reduces KV cache by 61% at 128K context length while maintaining performance comparable to full attention.
  arXiv  Detail & Related papers  (2025-12-10T01:54:57Z)
• Scaling LLM Speculative Decoding: Non-Autoregressive Forecasting in Large-Batch Scenarios [76.85739138203014]
  We present SpecFormer, a novel architecture that accelerates unidirectional and attention mechanisms.<n>We demonstrate that SpecFormer achieves lower training demands and reduced computational costs.
  arXiv  Detail & Related papers  (2025-11-25T14:20:08Z)
• Sparse Query Attention (SQA): A Computationally Efficient Attention Mechanism with Query Heads Reduction [0.0]
  This paper introduces Sparse Query Attention (SQA), a novel attention architecture that pursues an alternative and complementary optimization path.<n>It can achieve significant throughput improvements of up to 3x in computation-bound scenarios such as model pre-training, fine-tuning, and encoder-based tasks.<n>SQA was discovered serendipitously during the development of the upcoming Reactive Transformer architecture.
  arXiv  Detail & Related papers  (2025-10-02T09:01:38Z)
• Multipole Attention for Efficient Long Context Reasoning [64.94673641704289]
  Large Reasoning Models (LRMs) have shown promising accuracy improvements on complex problem-solving tasks.<n>LRMs need to generate long chain-of-thought reasoning in order to think before answering.<n>We introduce Multipole Attention, which accelerates autoregressive reasoning by only computing exact attention for the most important tokens.
  arXiv  Detail & Related papers  (2025-06-16T03:00:40Z)
• Curse of High Dimensionality Issue in Transformer for Long-context Modeling [31.257769500741006]
  We propose textitDynamic Group Attention (DGA) to reduce redundancy by aggregating less important tokens during attention computation.<n>Our results show that our DGA significantly reduces computational costs while maintaining competitive performance.
  arXiv  Detail & Related papers  (2025-05-28T08:34:46Z)
• Delta Attention: Fast and Accurate Sparse Attention Inference by Delta Correction [52.14200610448542]
  A transformer has a quadratic complexity, leading to high inference costs and latency for long sequences.<n>We propose a simple, novel, and effective procedure for correcting this distributional shift.<n>Our method can maintain approximately 98.5% sparsity over full quadratic attention, making our model 32 times faster than Flash Attention 2 when processing 1M token prefills.
  arXiv  Detail & Related papers  (2025-05-16T13:48:33Z)
• Task-Oriented Feature Compression for Multimodal Understanding via Device-Edge Co-Inference [54.53508601749513]
  We propose a task-oriented feature compression (TOFC) method for multimodal understanding in a device-edge co-inference framework.<n>To enhance compression efficiency, multiple entropy models are adaptively selected based on the characteristics of the visual features.<n>Results show that TOFC achieves up to 52% reduction in data transmission overhead and 63% reduction in system latency.
  arXiv  Detail & Related papers  (2025-03-17T08:37:22Z)
• Top-Theta Attention: Sparsifying Transformers by Compensated Thresholding [3.9826635165229223]
  We present Top-Theta (Top-$theta$) Attention, a training-free method for sparsifying transformer attention during inference.<n>Our key insight is that static, per-head thresholds can be calibrated to retain the desired constant number of significant elements per attention row.<n>Top-Theta$ achieves 3-10x reduction in V-cache usage and up to 10x fewer attention elements during inference while no more than 1% in accuracy.
  arXiv  Detail & Related papers  (2025-02-12T12:50:15Z)
• Identify Critical KV Cache in LLM Inference from an Output Perturbation Perspective [19.447729423696096]
  We propose a perturbation-constrained selection algorithm to identify critical KV cache entries.<n>Our algorithm achieves lower output perturbations in over 92% attention heads in Llama model.
  arXiv  Detail & Related papers  (2025-02-06T06:31:47Z)
• Anchor Attention, Small Cache: Code Generation with Large Language Models [15.94784908771546]
  Current practices in NLP often use sparse attention which may, unfortunately, lead to substantial inaccuracies, or hallucinations, in code generation tasks. We propose a novel approach, AnchorCoder, which features token-wise anchor attention designed to extract and compress contextual information. It can consistently achieve a significant (at least 70%) reduction in KV cache requirements, while preserving the majority of model's performance.
  arXiv  Detail & Related papers  (2024-11-11T02:47:05Z)
• ZipVL: Efficient Large Vision-Language Models with Dynamic Token Sparsification [29.163757099307553]
  The efficiency of large vision-language models (LVLMs) is constrained by the computational bottleneck of the attention mechanism during the prefill phase.<n>We present ZipVL, an efficient inference framework designed for LVLMs through a dynamic ratio allocation strategy of important tokens.
  arXiv  Detail & Related papers  (2024-10-11T07:24:21Z)
• ThinK: Thinner Key Cache by Query-Driven Pruning [63.13363917871414]
  Large Language Models (LLMs) have revolutionized the field of natural language processing, achieving unprecedented performance across a variety of applications.<n>This paper focuses on the long-context scenario, addressing the inefficiencies in KV cache memory consumption during inference.<n>We propose ThinK, a novel query-dependent KV cache pruning method designed to minimize attention weight loss while selectively pruning the least significant channels.
  arXiv  Detail & Related papers  (2024-07-30T17:59:08Z)
• ClusTR: Exploring Efficient Self-attention via Clustering for Vision Transformers [70.76313507550684]
  We propose a content-based sparse attention method, as an alternative to dense self-attention. Specifically, we cluster and then aggregate key and value tokens, as a content-based method of reducing the total token count. The resulting clustered-token sequence retains the semantic diversity of the original signal, but can be processed at a lower computational cost.
  arXiv  Detail & Related papers  (2022-08-28T04:18:27Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.