ET-Prune: Evidence-Aware Dynamic Budgeting for Visual Token Pruning in Text-Rich MLLMs

• URL: http://arxiv.org/abs/2608.01979v1
• Date: Mon, 03 Aug 2026 09:42:34 GMT
• Title: ET-Prune: Evidence-Aware Dynamic Budgeting for Visual Token Pruning in Text-Rich MLLMs
• Abstract summary: We present ET-Prune, a training-free framework that casts pruning as evidence allocation.<n>It derives question-conditioned evidence from a decoder-side partial query-key block.<n>It safeguards text-like spatial regions, and converts evidence uncertainty and density into a sample-specific token floor.
• Score: 37.20525993006875
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Visual token pruning reduces the inference cost of multimodal large language models, but a fixed token ratio is poorly matched to text-rich inputs. In OCR-centric tasks, decisive evidence can be a small number, label, or field whose relevance is specified by the question; indiscriminate pruning can erase that evidence while retaining visually salient but irrelevant regions. We present ET-Prune, a training-free framework that casts pruning as evidence allocation. It derives question-conditioned evidence from a decoder-side partial query-key block, safeguards text-like spatial regions, and converts evidence uncertainty and density into a sample-specific token floor. Three progressive middle-layer events then move the sequence toward this budget, retaining more tokens for diffuse or text-dense evidence and pruning concentrated evidence more aggressively. At the observed point estimates from one deterministic pass per configuration, ET-Prune leads or ties among pruned methods in all six backbone-benchmark comparisons at roughly half tokens. On OCRBench-v2, it leads the strongest pruned baselines by 1.80 and 0.68 percentage points on Qwen3-VL-8B and InternVL3.5-8B, respectively, while retaining about half of the visual tokens; on MMBench v1.1, it reaches 0.8467 circular exact-matching accuracy versus 0.8437 for Vanilla at 54.45% average visual-token retention. These results show a favorable observed quality-cost trade-off for evidence-aware dynamic budgeting in text-rich multimodal inference.

Related papers

• E2S-Pruner: Progressive Two-Stage Evidence Fusion for Visual Token Pruning in Vision-Language Models [12.765837057758127]
  We propose a two-stage evidence-fusion framework for visual token pruning.<n>E2S-Pruner represents each visual token using three states: important, unimportant, and uncertain.<n>On LLaVA-1.5-7B, E2S-Pruner retains 98.0%, 96.8%, and 90.6% of the aggregate performance when the average numbers of retained visual tokens are 192, 128, and 64.
  arXiv  Detail & Related papers  (2026-08-24T13:44:41Z)
• RoRA: Role-Oriented Regional Allocation for Visual Token Pruning in MLLMs [50.93650157115467]
  We present RoRA, a training-free framework that casts visual token pruning as role-oriented regional evidence allocation.<n>Under matched budgets, RoRA consistently outperforms strong training-free baselines across LLaVA and Qwen-VL families.<n>At a 66.7% pruning ratio, RoRA requires only 0.7 ms for token selection and reduces end-to-end inference time by 24.6%.
  arXiv  Detail & Related papers  (2026-08-07T10:39:47Z)
• DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models [13.866355425347384]
  Visual inputs in vision-language models are often encoded into substantially longer token sequences than text.<n>Recent methods address this bottleneck by scoring token importance and pruning low-scoring tokens in a single pass.<n>Motivated by this insight, we introduce DIVE, a training-free framework that recasts visual-token pruning as dynamic evidence construction.
  arXiv  Detail & Related papers  (2026-08-05T06:32:25Z)
• Beyond Accuracy: Auditing Spatial Provenance in Visual Token Pruning for OCR-Critical MLLM Inference [22.335515110971475]
  Visual-token pruning is judged by answer quality at a fixed retention budget.<n>We turn this blind spot into an evidence-risk audit that couples answer behavior with geometric token-origin provenance.<n>We show that favorable target-verification points do not imply task-general compression.
  arXiv  Detail & Related papers  (2026-07-29T13:45:31Z)
• SepPrune:A Separator-based Pruning Framework for Efficient Multimodal Large Language Models [22.517061458809057]
  We propose SepPrune, an efficient, training-free, plug-and-play pruning method for vision tokens.<n>SepPrune reuses the LLM's built-in projection parameters and requires no architectural changes.<n> Experiments on Qwen2.5-VL-7B show that SepPrune achieves state-of-the-art performance, retaining 96.3% of the original accuracy while removing 80.2% of vision tokens.
  arXiv  Detail & Related papers  (2026-07-28T15:01:02Z)
• OccamToken: Efficient VLM Inference with Training-Free and Budget-Adaptive Token Pruning [24.164883144694656]
  OccamToken is a training-free framework that replaces absolute token ranking with register-anchored relative evidence testing.<n>We show that OccamToken consistently improves the accuracy-efficiency trade-off without additional training.
  arXiv  Detail & Related papers  (2026-05-28T09:20:47Z)
• Progressive Online Video Understanding with Evidence-Aligned Timing and Transparent Decisions [75.23170605943457]
  textbfmodel is a framework that decouples reasoning control from memory integration.<n>emphActive Thinking Decision Maker (ATDM) is a transparent reasoning controller that externalizes its decision process.<n>emphHierarchical Progressive Semantic Integration (HPSI) module acts as an efficient memory system.
  arXiv  Detail & Related papers  (2026-04-20T16:15:33Z)
• Prune Redundancy, Preserve Essence: Vision Token Compression in VLMs via Synergistic Importance-Diversity [54.95089105944234]
  Vision-language models (VLMs) face significant computational inefficiencies caused by excessive generation of visual tokens.<n>We propose PruneSID, a training-free Synergistic Importance-Diversity approach featuring a two-stage pipeline.<n>We demonstrate state-of-the-art performance, achieving 96.3% accuracy on LLaVA-1.5 with only 11.1% token retention, and 92.8% accuracy at extreme compression rates (5.6%) on LLaVA-NeXT.
  arXiv  Detail & Related papers  (2026-03-10T10:31:58Z)
• Focus-Scan-Refine: From Human Visual Perception to Efficient Visual Token Pruning [78.75062483648243]
  Vision-language models (VLMs) often generate massive visual tokens that greatly increase inference latency and memory footprint.<n>We propose Focus-Scan-Refine (FSR), a human-inspired, plug-and-play pruning framework that mimics how humans answer visual questions.<n>FSR consistently improves the accuracy-efficiency trade-off over existing state-of-the-art pruning methods.
  arXiv  Detail & Related papers  (2026-02-05T16:02:48Z)
• Look As You Think: Unifying Reasoning and Visual Evidence Attribution for Verifiable Document RAG via Reinforcement Learning [55.232400251303794]
  Look As You Think (LAT) is a reinforcement learning framework that trains models to produce verifiable reasoning paths with consistent attribution.<n>LAT consistently improves the vanilla model in both single- and multi-image settings, yielding average gains of 8.23% in soft exact match (EM) and 47.0% in IoU@0.5.
  arXiv  Detail & Related papers  (2025-11-15T02:50:23Z)

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.