Leveraging KV Similarity for Online Structured Pruning in LLMs

• URL: http://arxiv.org/abs/2512.07090v1
• Date: Mon, 08 Dec 2025 01:56:27 GMT
• Title: Leveraging KV Similarity for Online Structured Pruning in LLMs
• Authors: Jungmin Lee, Gwangeun Byeon, Yulhwa Kim, Seokin Hong,
• Abstract summary: Pruning has emerged as a promising direction for accelerating large language model (LLM) inference.<n>We introduce Token Filtering, a lightweight online structured pruning technique that makes pruning decisions directly during inference without any calibration data.<n>Tests on LLaMA-2 (7B/13B), LLaMA-3 (8B), and Mistral (7B) demonstrate that Token Filtering consistently outperforms prior structured pruning methods.
• Score: 7.067238287331514
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Pruning has emerged as a promising direction for accelerating large language model (LLM) inference, yet existing approaches often suffer from instability because they rely on offline calibration data that may not generalize across inputs. In this work, we introduce Token Filtering, a lightweight online structured pruning technique that makes pruning decisions directly during inference without any calibration data. The key idea is to measure token redundancy via joint key-value similarity and skip redundant attention computations, thereby reducing inference cost while preserving critical information. To further enhance stability, we design a variance-aware fusion strategy that adaptively weights key and value similarity across heads, ensuring that informative tokens are retained even under high pruning ratios. This design introduces no additional memory overhead and provides a more reliable criterion for token importance. Extensive experiments on LLaMA-2 (7B/13B), LLaMA-3 (8B), and Mistral (7B) demonstrate that Token Filtering consistently outperforms prior structured pruning methods, preserving accuracy on commonsense reasoning benchmarks and maintaining strong performance on challenging tasks such as MMLU, even with 50% pruning.

Related papers

• Thinking by Subtraction: Confidence-Driven Contrastive Decoding for LLM Reasoning [58.331709210563616]
  Thinking by Subtraction is a confidence-driven contrastive decoding approach.<n>A small subset of low-confidence tokens disproportionately contributes to reasoning errors and unnecessary output expansion.<n>Our method, Confidence-Driven Contrastive Decoding, detects low-confidence tokens during decoding and intervenes at these positions.
  arXiv  Detail & Related papers  (2026-02-20T14:13:22Z)
• CORE: Context-Robust Remasking for Diffusion Language Models [51.59514489363897]
  We propose Context-Robust Remasking (CORE), a training-free framework for inference-time revision.<n>Rather than trusting static token probabilities, CORE identifies context-brittle tokens by probing their sensitivity to targeted masked-context perturbations.<n>On LLaDA-8B-Base, CORE delivers consistent improvements across reasoning and code benchmarks, outperforming compute-matched baselines and improving MBPP by up to 9.2 percentage points.
  arXiv  Detail & Related papers  (2026-02-04T00:12:30Z)
• Learn More with Less: Uncertainty Consistency Guided Query Selection for RLVR [18.494852448006462]
  Existing RLVR algorithms require large query budgets, making annotation costly.<n>We investigate whether fewer but more informative queries can yield similar or superior performance, introducing active learning (AL) into RLVR.<n>Experiments show our method consistently outperforms random and classic AL baselines, achieving full-dataset performance while training on only 30% of the data.
  arXiv  Detail & Related papers  (2026-01-30T05:41:55Z)
• D2Pruner: Debiased Importance and Structural Diversity for MLLM Token Pruning [49.16227597771663]
  D2Pruner is a framework that combines debiased importance with a structural pruning mechanism.<n>It reduces FLOPs by 74.2% while retaining 99.2% of its original performance.<n>It marks a significant advancement with up to 63. 53% improvement over existing methods.
  arXiv  Detail & Related papers  (2025-12-22T14:42:31Z)
• 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)
• Winning the Pruning Gamble: A Unified Approach to Joint Sample and Token Pruning for Efficient Supervised Fine-Tuning [71.30276778807068]
  We propose a unified framework that strategically coordinates sample pruning and token pruning.<n>Q-Tuning achieves a +38% average improvement over the full-data SFT baseline using only 12.5% of the original training data.
  arXiv  Detail & Related papers  (2025-09-28T13:27:38Z)
• ERIS: An Energy-Guided Feature Disentanglement Framework for Out-of-Distribution Time Series Classification [51.07970070817353]
  An ideal time series classification (TSC) should be able to capture invariant representations.<n>Current methods are largely unguided, lacking the semantic direction required to isolate truly universal features.<n>We propose an end-to-end Energy-Regularized Information for Shift-Robustness framework to enable guided and reliable feature disentanglement.
  arXiv  Detail & Related papers  (2025-08-19T12:13:41Z)
• Confident or Seek Stronger: Exploring Uncertainty-Based On-device LLM Routing From Benchmarking to Generalization [61.02719787737867]
  Large language models (LLMs) are increasingly deployed and democratized on edge devices.<n>One promising solution is uncertainty-based SLM routing, offloading high-stakes queries to stronger LLMs when resulting in low-confidence responses on SLM.<n>We conduct a comprehensive investigation into benchmarking and generalization of uncertainty-driven routing strategies from SLMs to LLMs over 1500+ settings.
  arXiv  Detail & Related papers  (2025-02-06T18:59:11Z)
• How Breakable Is Privacy: Probing and Resisting Model Inversion Attacks in Collaborative Inference [13.453033795109155]
  Collaborative inference improves computational efficiency for edge devices by transmitting intermediate features to cloud models.<n>There is no established criterion for assessing the difficulty of model inversion attacks (MIAs)<n>We propose the first theoretical criterion to assess MIA difficulty in CI, identifying mutual information, entropy, and effective information volume as key influencing factors.
  arXiv  Detail & Related papers  (2025-01-01T13:00:01Z)
• More Tokens, Lower Precision: Towards the Optimal Token-Precision Trade-off in KV Cache Compression [71.42818367729573]
  In large language models (LLMs), the memory usage of KV cache has become a critical bottleneck during inference.<n>The mainstream KV compression methods, including KV pruning and KV quantization, primarily focus on either token or precision dimension separately.<n>In this paper, we comprehensively investigate the token-precision trade-off in KV cache compression.
  arXiv  Detail & Related papers  (2024-12-17T09:20:31Z)
• LLM-BIP: Structured Pruning for Large Language Models with Block-Wise Forward Importance Propagation [0.0]
  We propose a more accurate pruning metric based on the block-wise importance score propagation.<n>We evaluate the proposed method using LLaMA-7B, Vicuna-7B, and LLaMA-13B across common zero-shot tasks.
  arXiv  Detail & Related papers  (2024-12-09T11:57:16Z)
• Reliability-Adaptive Consistency Regularization for Weakly-Supervised Point Cloud Segmentation [80.07161039753043]
  Weakly-supervised point cloud segmentation with extremely limited labels is desirable to alleviate the expensive costs of collecting densely annotated 3D points. This paper explores applying the consistency regularization that is commonly used in weakly-supervised learning, for its point cloud counterpart with multiple data-specific augmentations. We propose a novel Reliability-Adaptive Consistency Network (RAC-Net) to use both prediction confidence and model uncertainty to measure the reliability of pseudo labels.
  arXiv  Detail & Related papers  (2023-03-09T10:41:57Z)

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.