SepLLM: Accelerate Large Language Models by Compressing One Segment into One Separator

• URL: http://arxiv.org/abs/2412.12094v6
• Date: Mon, 02 Jun 2025 11:46:43 GMT
• Title: SepLLM: Accelerate Large Language Models by Compressing One Segment into One Separator
• Abstract summary: Large Language Models (LLMs) have exhibited exceptional performance across a spectrum of natural language processing tasks.<n>We have identified a key pattern: certain seemingly meaningless separator tokens (i.e., punctuations) contribute disproportionately to attention scores compared to semantically meaningful tokens.<n>We introduce SepLLM, a plug-and-play framework that accelerates inference by compressing these segments and eliminating redundant tokens.
• Score: 65.62084602011596
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Language Models (LLMs) have exhibited exceptional performance across a spectrum of natural language processing tasks. However, their substantial sizes pose considerable challenges, particularly in computational demands and inference speed, due to their quadratic complexity. In this work, we have identified a key pattern: certain seemingly meaningless separator tokens (i.e., punctuations) contribute disproportionately to attention scores compared to semantically meaningful tokens. This observation suggests that information of the segments between these separator tokens can be effectively condensed into the separator tokens themselves without significant information loss. Guided by this insight, we introduce SepLLM, a plug-and-play framework that accelerates inference by compressing these segments and eliminating redundant tokens. Additionally, we implement efficient kernels for training acceleration. Experimental results across training-free, training-from-scratch, and post-training settings demonstrate SepLLM's effectiveness. Notably, using the Llama-3-8B backbone, SepLLM achieves over 50% reduction in KV cache on the GSM8K-CoT benchmark while maintaining comparable performance. Furthermore, in streaming settings, SepLLM effectively processes sequences of up to 4 million tokens or more while maintaining consistent language modeling capabilities.

Related papers

• Beyond Uniform Tokens: Adaptive Compression for Time Series Language Models [5.75338138628308]
  We study token efficiency in time series language modeling from an asymmetric-token perspective.<n>We show that TS tokens have highly uneven spectral contributions, where many tokens share redundant frequency patterns while a small subset preserves critical temporal evidence.<n>We develop an adaptive token budgeting framework that compresses TS tokens via frequency-domain structure and progressively reduces prompt tokens across layers.
  arXiv  Detail & Related papers  (2026-06-11T17:39:26Z)
• Grammatically-Guided Sparse Attention for Efficient and Interpretable Transformers [0.0]
  Grammatically-Guided Sparse Attention constrains attention computations based on the grammatical roles of tokens.<n>Two masking strategies are proposed and evaluated.<n>Preliminary results show accuracy values of 0.8200 for hard masking and 0.8165 for soft masking.
  arXiv  Detail & Related papers  (2026-05-23T11:11:19Z)
• EvoComp: Learning Visual Token Compression for Multimodal Large Language Models via Semantic-Guided Evolutionary Labeling [7.872065731262025]
  EvoComp is a visual token compression framework that significantly reduces token count while preserving task accuracy.<n>We show that it retains 99.3% of the original accuracy under 3x token compression and delivers up to 1.6x speedup on mobile devices.
  arXiv  Detail & Related papers  (2026-04-18T17:52:02Z)
• SepSeq: A Training-Free Framework for Long Numerical Sequence Processing in LLMs [51.84813231879529]
  We propose a training-free, plug-and-play framework to mitigate dispersion by strategically inserting separator tokens.<n>Mechanistically, we demonstrate that separator tokens act as an attention sink, recalibrating attention to focus on local segments while preserving global context.
  arXiv  Detail & Related papers  (2026-04-09T02:40:34Z)
• Beyond Attention Magnitude: Leveraging Inter-layer Rank Consistency for Efficient Vision-Language-Action Models [75.27089289058613]
  We introduce textbfTIES (textbfTau-guided textbfInter-layer textbfEfficient textbfSelection), a dynamic framework guided by inter-layer token ranking consistency.<n>On the CogACT + SIMPLER benchmark, TIES improves average success rates by 6% while reducing token usage by 78%, and demonstrate strong generalization across diverse decoders and benchmarks.
  arXiv  Detail & Related papers  (2026-03-26T02:13:03Z)
• DyLLM: Efficient Diffusion LLM Inference via Saliency-based Token Selection and Partial Attention [2.7422645382944935]
  We present DyLLM, a training-free inference framework that accelerates decoding by selectively computing only salient tokens.<n>Across diverse reasoning and code-generation benchmarks, DyLLM achieves up to 9.6x higher throughput.
  arXiv  Detail & Related papers  (2026-03-09T07:02:01Z)
• LiteToken: Removing Intermediate Merge Residues From BPE Tokenizers [76.59130257385826]
  Intermediate merge residues in BPE vocabularies are frequent during merge learning so that retained in the final vocabulary, but are mostly further merged and rarely emitted when tokenizing the corpus during tokenizer usage.<n>We present a systematic empirical characterization of this phenomenon across commonly used tokenizers and introduce LiteToken, a simple method for removing residue tokens.<n>Experiments show that LiteToken reduces token fragmentation, reduces parameters, and improves robustness to noisy or misspelled inputs, while preserving overall performance.
  arXiv  Detail & Related papers  (2026-02-04T16:19:05Z)
• Fast SAM2 with Text-Driven Token Pruning [52.8350457627401]
  Segment Anything Model 2 (SAM2), a vision computation model has significantly advanced in prompt-driven video object segmentation.<n>SAM2 pipelines propagate all visual tokens produced by the image encoder through downstream temporal reasoning modules, regardless of their relevance to the target object.<n>We introduce a text-guided token pruning framework that improves inference efficiency by selectively reducing token density prior to temporal propagation.
  arXiv  Detail & Related papers  (2025-12-24T18:59:05Z)
• TrimTokenator: Towards Adaptive Visual Token Pruning for Large Multimodal Models [4.779482139419908]
  We introduce a mutual information-based token pruning strategy that removes visual tokens semantically with textual tokens.<n>Our method maintains strong performance while reducing textual tokens by 88.9% on models such as LLaVA-15-7B and LLaVA--7B.
  arXiv  Detail & Related papers  (2025-08-30T02:43:50Z)
• 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)
• PACT: Pruning and Clustering-Based Token Reduction for Faster Visual Language Models [0.0]
  We introduce PACT, a method that reduces inference time and memory usage by pruning irrelevant tokens and merging visually redundant ones. Our approach uses a novel importance metric to identify unimportant tokens without relying on attention scores. We also propose a novel clustering algorithm, called Distance Bounded Density Peak Clustering, which efficiently clusters visual tokens.
  arXiv  Detail & Related papers  (2025-04-11T20:45:00Z)
• SentenceKV: Efficient LLM Inference via Sentence-Level Semantic KV Caching [9.617322424513317]
  SentenceKV is a novel KV caching approach designed to enhance inference efficiency while preserving semantic coherence. We show that SentenceKV significantly outperforms state-of-the-art methods in both efficiency and memory usage, without compromising model accuracy.
  arXiv  Detail & Related papers  (2025-04-01T17:08:57Z)
• Compressing KV Cache for Long-Context LLM Inference with Inter-Layer Attention Similarity [24.118503938098307]
  Existing methods, including selective token retention and window-based attention, improve efficiency but risk discarding important tokens needed for future text generation.<n>We propose an approach that enhances LLM efficiency without token loss by reducing the memory and computational load of less important tokens, rather than discarding them.
  arXiv  Detail & Related papers  (2024-12-03T08:29:27Z)
• Squeezed Attention: Accelerating Long Context Length LLM Inference [64.11145320159126]
  We propose Squeezed Attention as a mechanism to accelerate LLM applications where a large portion of the input prompt is fixed. We use K-means clustering offline to group the keys for the fixed context based on semantic similarity and represent each cluster with a single centroid value. We then compute exact attention using only these important keys from the fixed context, thereby reducing bandwidth and computational costs.
  arXiv  Detail & Related papers  (2024-11-14T18:54:19Z)
• MrT5: Dynamic Token Merging for Efficient Byte-level Language Models [50.46453950887946]
  This work introduces MrT5 (MergeT5), a more efficient variant of ByT5. MrT5 integrates a token deletion mechanism in its encoder to dynamically shorten the input sequence length. When trained on English text, MrT5 demonstrates the capability to transfer its deletion feature zero-shot across several languages.
  arXiv  Detail & Related papers  (2024-10-28T06:14:12Z)
• Semantic Equitable Clustering: A Simple and Effective Strategy for Clustering Vision Tokens [57.37893387775829]
  We introduce a fast and balanced clustering method, named textbfSemantic textbfEquitable textbfClustering (SEC) SEC clusters tokens based on their global semantic relevance in an efficient, straightforward manner. We propose a versatile vision backbone, SECViT, to serve as a vision language connector.
  arXiv  Detail & Related papers  (2024-05-22T04:49:00Z)
• Training-Free Long-Context Scaling of Large Language Models [114.53296002607993]
  We propose Dual Chunk Attention, which enables Llama2 70B to support context windows of more than 100k tokens without continual training. By decomposing the attention for long sequences into chunk-based modules, DCA manages to effectively capture the relative positional information of tokens.
  arXiv  Detail & Related papers  (2024-02-27T12:39:23Z)
• MST: Adaptive Multi-Scale Tokens Guided Interactive Segmentation [8.46894039954642]
  We propose a novel multi-scale token adaptation algorithm for interactive segmentation. By performing top-k operations across multi-scale tokens, the computational complexity is greatly simplified. We also propose a token learning algorithm based on contrastive loss.
  arXiv  Detail & Related papers  (2024-01-09T07:59:42Z)
• AiluRus: A Scalable ViT Framework for Dense Prediction [95.1313839257891]
  Vision transformers (ViTs) have emerged as a prevalent architecture for vision tasks owing to their impressive performance. We propose to apply adaptive resolution for different regions in the image according to their importance. We evaluate our proposed method on three different datasets and observe promising performance.
  arXiv  Detail & Related papers  (2023-11-02T12:48:43Z)
• UNETR++: Delving into Efficient and Accurate 3D Medical Image Segmentation [93.88170217725805]
  We propose a 3D medical image segmentation approach, named UNETR++, that offers both high-quality segmentation masks as well as efficiency in terms of parameters, compute cost, and inference speed. The core of our design is the introduction of a novel efficient paired attention (EPA) block that efficiently learns spatial and channel-wise discriminative features. Our evaluations on five benchmarks, Synapse, BTCV, ACDC, BRaTs, and Decathlon-Lung, reveal the effectiveness of our contributions in terms of both efficiency and accuracy.
  arXiv  Detail & Related papers  (2022-12-08T18:59:57Z)

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.