Language-Guided Token Compression with Reinforcement Learning in Large Vision-Language Models

• URL: http://arxiv.org/abs/2603.13394v1
• Date: Wed, 11 Mar 2026 09:52:44 GMT
• Title: Language-Guided Token Compression with Reinforcement Learning in Large Vision-Language Models
• Abstract summary: TPRL learns adaptive pruning trajectories through language-guided sequential optimization tied directly to end-task performance.<n>We formulate visual token pruning as a sequential decision process with explicit state transitions.<n>We employ a self-supervised autoencoder to compress visual tokens into a compact state representation for efficient policy learning.
• Score: 19.87835200940665
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Vision-Language Models (LVLMs) incur substantial inference costs due to the processing of a vast number of visual tokens. Existing methods typically struggle to model progressive visual token reduction as a multi-step decision process with sequential dependencies and often rely on hand-engineered scoring rules that lack adaptive optimization for complex reasoning trajectories. To overcome these limitations, we propose TPRL, a reinforcement learning framework that learns adaptive pruning trajectories through language-guided sequential optimization tied directly to end-task performance. We formulate visual token pruning as a sequential decision process with explicit state transitions and employ a self-supervised autoencoder to compress visual tokens into a compact state representation for efficient policy learning. The pruning policy is initialized through learning from demonstrations and subsequently fine-tuned using Proximal Policy Optimization (PPO) to jointly optimize task accuracy and computational efficiency. Our experimental results demonstrate that TPRL removes up to 66.7\% of visual tokens and achieves up to a 54.2\% reduction in FLOPs during inference while maintaining a near-lossless average accuracy drop of only 0.7\%. Code is released at \href{https://github.com/MagicVicCoder/TPRL}{\textcolor{mypink}{https://github.com/MagicVicCoder/TPRL}}.

Related papers

• RCP: Representation Consistency Pruner for Mitigating Distribution Shift in Large Vision-Language Models [24.3914653184824]
  Large Vision-Language Models (LVLMs) suffer from prohibitive inference costs due to the massive number of visual tokens processed by the language decoder.<n>Existing pruning methods often lead to significant performance degradation because the irreversible removal of visual tokens causes a distribution shift in the hidden states that deviates from the pre-trained full-token regime.<n>We propose Consistency Representation Pruner, which integrates cumulative visual token pruning with a delayed repair mechanism.
  arXiv  Detail & Related papers  (2026-04-04T13:31:45Z)
• Not All Tokens See Equally: Perception-Grounded Policy Optimization for Large Vision-Language Models [38.47027398567909]
  Perception-Grounded Policy Optimization (PGPO) is a novel fine-grained credit assignment framework that dynamically reshapes advantages at the token level.<n>We show that PGPO actively amplifies learning signals for visually-dependent tokens while suppressing gradient noise from linguistic priors.<n>Both theoretical and empirical analyses confirm that PGPO effectively reduces gradient variance, prevents training collapse, and acts as a potent regularizer for robust, perception-grounded multimodal reasoning.
  arXiv  Detail & Related papers  (2026-04-02T09:53:20Z)
• Rethinking Token Reduction for Large Vision-Language Models [95.48478689025696]
  Large-Language Models (LVLMs) excel in visual understanding and reasoning, but the excessive visual tokens lead to high inference costs.<n>We propose a learning-based prompt-agnostic, termed MetaCompress, overcoming the limitations of generalization designs.
  arXiv  Detail & Related papers  (2026-03-23T08:40:08Z)
• ResPrune: Text-Conditioned Subspace Reconstruction for Visual Token Pruning in Large Vision-Language Models [59.94664910790462]
  ResPrune is a training-free visual token pruning framework for large vision-language models.<n>It selects a compact yet informative subset of visual tokens.<n>It achieves effective reductions in computation, memory consumption, and inference latency.
  arXiv  Detail & Related papers  (2026-03-22T07:44:45Z)
• ForgeryVCR: Visual-Centric Reasoning via Efficient Forensic Tools in MLLMs for Image Forgery Detection and Localization [62.03035862528452]
  ForgeryVCR is a framework that materializes imperceptible traces into explicit visual intermediates via Visual-Centric Reasoning.<n>ForgeryVCR achieves state-of-the-art (SOTA) performance in both detection and localization tasks.
  arXiv  Detail & Related papers  (2026-02-15T11:14:47Z)
• 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)
• Informed Routing in LLMs: Smarter Token-Level Computation for Faster Inference [7.690958366125321]
  This paper introduces informed routing, a new paradigm that proactively addresses these issues.<n>We propose the Lightweight Feature Forecaster (LFF), a small predictive module that estimates a unit's output before routing decisions are made.<n>Experiments on both language modeling and reasoning tasks show that informed routing achieves state-of-the-art efficiency-performance trade-offs.
  arXiv  Detail & Related papers  (2025-10-10T09:59:36Z)
• AutoPrune: Each Complexity Deserves a Pruning Policy [58.448785378705566]
  Complexity Pruning (AutoPrune) is a training-free, plug-and-play framework that tailors pruning policies to varying sample and task complexities.<n>We evaluate AutoPrune on standard vision-Adaptive tasks and on Vision-Language-Action models for autonomous driving.
  arXiv  Detail & Related papers  (2025-09-28T15:09:00Z)
• A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models [94.49953824684853]
  We introduce a dynamic pruning framework, GlimpsePrune, inspired by human cognition.<n>It takes a data-driven ''glimpse'' and prunes irrelevant visual tokens in a single forward pass before answer generation.<n>An enhanced GlimpsePrune+ achieves 110% of the baseline performance while maintaining a similarly high pruning rate.
  arXiv  Detail & Related papers  (2025-08-03T02:15:43Z)
• Skip-Vision: Efficient and Scalable Acceleration of Vision-Language Models via Adaptive Token Skipping [13.846838416902575]
  A key bottleneck stems from the proliferation of visual tokens required for fine-grained image understanding.<n>We propose Skip-Vision, a unified framework addressing both training and inference inefficiencies in vision-language models.<n> Experimental results demonstrate that Skip-Vision reduces training time by up to 35%, inference FLOPs by 75%, and latency by 45%.
  arXiv  Detail & Related papers  (2025-03-26T04:16:48Z)
• PRISM: Self-Pruning Intrinsic Selection Method for Training-Free Multimodal Data Selection [68.8373788348678]
  Visual instruction tuning adapts pre-trained Multimodal Large Language Models to follow human instructions.<n>PRISM is the first training-free framework for efficient visual instruction selection.<n>It reduces the end-to-end time for data selection and model tuning to just 30% of conventional pipelines.
  arXiv  Detail & Related papers  (2025-02-17T18:43:41Z)
• Pruning All-Rounder: Rethinking and Improving Inference Efficiency for Large Vision Language Models [42.124670377223175]
  We propose a novel framework for inference acceleration called the Pruning All-Rounder (PAR)<n>With a self-supervised learning manner, PAR achieves a superior balance between performance and efficiency. Notably, PAR is highly flexible, offering multiple pruning versions to address a range of acceleration scenarios.
  arXiv  Detail & Related papers  (2024-12-09T13:02:35Z)
• When Does Visual Prompting Outperform Linear Probing for Vision-Language Models? A Likelihood Perspective [57.05315507519704]
  We propose a log-likelihood ratio (LLR) approach to analyze the comparative benefits of visual prompting and linear probing. Our measure attains up to a 100-fold reduction in run time compared to full training, while achieving prediction accuracies up to 91%.
  arXiv  Detail & Related papers  (2024-09-03T12:03:45Z)

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.