Slow-Fast Inference: Training-Free Inference Acceleration via Within-Sentence Support Stability

• URL: http://arxiv.org/abs/2603.12038v1
• Date: Thu, 12 Mar 2026 15:14:48 GMT
• Title: Slow-Fast Inference: Training-Free Inference Acceleration via Within-Sentence Support Stability
• Abstract summary: We propose a training-free decoding framework that decouples generation into frequent low-cost fast steps and occasional dense-attention slow steps.<n>Across the evaluated context lengths, SFI delivers approximately $1.6times$--$14.4times$ higher decoding throughput.
• Score: 61.28049899686214
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Long-context autoregressive decoding remains expensive because each decoding step must repeatedly process a growing history. We observe a consistent pattern during decoding: within a sentence, and more generally within a short semantically coherent span, the dominant attention support often remains largely stable. Motivated by this observation, we propose Slow-Fast Inference (SFI), a training-free decoding framework that decouples generation into frequent low-cost fast steps and occasional dense-attention slow steps. Fast steps reuse a compact sparse memory for efficient decoding. Slow steps are triggered near semantic boundaries. At slow steps, the model revisits the broader context and uses the Selector to refresh the selected memory for subsequent fast steps. Across the evaluated context lengths, SFI delivers approximately $1.6\times$--$14.4\times$ higher decoding throughput while generally maintaining quality on par with the full-KV baseline across long-context and long-CoT settings. Because SFI is training-free and applies directly to existing checkpoints, it offers a practical path to reducing inference cost for contemporary autoregressive reasoning models in long-context, long-horizon, and agentic workloads.

Related papers

• CateKV: On Sequential Consistency for Long-Context LLM Inference Acceleration [67.17709877327232]
  Cate KV is a hybrid KV cache method that retains only critical token information for consistent heads.<n>Cate KV reduces memory usage by up to $2.72times$ and accelerates decoding by $2.18times$ in single-sample inputs.
  arXiv  Detail & Related papers  (2026-08-31T06:02:37Z)
• You Only Index Once: Cross-Layer Sparse Attention with Shared Routing [61.29627714699688]
  Cross-layer sparse attention (A) is built on top of KV-sharing architectures such as YOCO.<n>A single indexer computes token-level top-k selection once and reuses the resulting index across layers.<n>Results show CLSA is both accurate and efficient, achieving up to 7.6x decoding speedup and 17.1x overall throughput improvement at 128K context.
  arXiv  Detail & Related papers  (2026-06-04T17:54:04Z)
• Kwai Summary Attention Technical Report [69.40814939510126]
  Long-context ability, has become one of the most important direction of next-generation Large Language Models.<n>Standard softmax attention exhibits quadratic time complexity with respect to sequence length.<n>We propose Kwai Summary Attention (KSA), a novel attention mechanism that reduces sequence modeling cost by compressing historical contexts.
  arXiv  Detail & Related papers  (2026-04-27T12:59:53Z)
• CSAttention: Centroid-Scoring Attention for Accelerating LLM Inference [3.1255988998610307]
  Centroid-Scoring Attention (CSAttention) is a training-free sparse attention method optimized for high- throughput contexts.<n>It front-loads computation into a one-time offline prefill phase that can be amortized across multiple queries.<n>It consistently outperforms state-of-the-art sparse attention methods in both model accuracy and inference speed.
  arXiv  Detail & Related papers  (2026-03-30T01:42:29Z)
• 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)
• Retrospective Sparse Attention for Efficient Long-Context Generation [5.562294018150909]
  RetroAttention retrospectively revises past attention outputs using newly arrived KV entries from subsequent decoding steps.<n>This breaks the fixed-attention-output paradigm and allows continual correction of prior approximations.<n>Experiments show that RetroAttention consistently outperforms state-of-the-art (SOTA) KV compression methods.
  arXiv  Detail & Related papers  (2025-08-12T15:11:47Z)
• Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache Eviction [72.27673320976933]
  Diffusion Large Language Models (dLLMs) enable breakthroughs in reasoning and parallel decoding.<n>Current caching techniques accelerate decoding by storing full-layer states, yet impose substantial memory usage.<n>We propose Sparse-dLLM, the first training-free framework integrating dynamic cache eviction with sparse attention.
  arXiv  Detail & Related papers  (2025-08-04T16:14:03Z)
• R-Stitch: Dynamic Trajectory Stitching for Efficient Reasoning [80.104336426172]
  Chain-of-thought (CoT) enhances problem-solving ability of large language models.<n>CoT incurs substantial inference cost due to long autoregressive trajectories.<n>We introduce R-Stitch, a training-free hybrid decoding framework.
  arXiv  Detail & Related papers  (2025-07-23T08:14:36Z)
• Rectified Sparse Attention [61.7702154360081]
  Efficient long-sequence generation is a critical challenge for Large Language Models.<n>We propose Rectified Sparse Attention (ReSA), a simple yet effective method that combines block-sparse attention with periodic dense rectification.<n> Experiments across math reasoning, language modeling, and retrieval tasks demonstrate that ReSA achieves near-lossless generation quality.
  arXiv  Detail & Related papers  (2025-06-04T16:01:48Z)
• SpecExtend: A Drop-in Enhancement for Speculative Decoding of Long Sequences [11.225649178057695]
  SpecExtend improves speculative decoding on long sequences without additional training.<n>To improve both draft accuracy and speed on long inputs without retraining, we propose Cross-model Retrieval.<n>SpecExtend accelerates speculative decoding by up to 2.84x on 16K-token long summarization and up to 3.86x on long reasoning.
  arXiv  Detail & Related papers  (2025-05-27T06:30:00Z)
• Dialogue Without Limits: Constant-Sized KV Caches for Extended Responses in LLMs [6.222287867011644]
  We propose MorphKV, an inference-time technique that maintains a constant-sized KV cache while preserving accuracy.<n>Unlike retention or lossy compression, MorphKV iteratively refines the KV cache via lightweight updates guided by attention patterns of recent tokens.<n>Our studies show 52.9$%$ memory savings and 18.2$%$ higher accuracy on average compared to state-of-the-art prior works.
  arXiv  Detail & Related papers  (2025-03-02T18:12:50Z)

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.