Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Evading Visual Aphasia: Contrastive Adaptive Semantic Token Pruning for Vision-Language Models

About

Are low-attention visual tokens truly redundant in vision-language reasoning? Existing pruning methods often assume so, ranking visual tokens by shallow text-to-image attention and discarding low-scoring patches to accelerate LVLM inference. We show that this scalar criterion is unreliable for compositional reasoning: tokens ignored in early layers can later become essential for resolving secondary objects, spatial relations, and contextual cues. Premature pruning can therefore induce Visual Aphasia, a failure mode in which the model loses visual grounding and falls back on language priors. We introduce COAST (COntrastive Adaptive Semantic Token Pruning), a training-free pruning framework that casts compression as adaptive semantic routing. COAST uses native cross-modal attention to identify query-specific anchors and estimate contextual dispersion via attention entropy, then adapts the retention trade-off between semantic evidence and spatial context. It further uses a contrastive routing score to preserve both anchor-aligned evidence and complementary spatial context. Across seven benchmarks, COAST reduces visual tokens by 77.8% and achieves a 2.15x latency speedup while retaining 98.64% of the original average performance. Beyond a single backbone or compression setting, COAST consistently outperforms strong pruning baselines across token budgets and generalizes across multiple LVLM families, showing that adaptive semantic routing is a robust alternative to one-shot scalar pruning

Jie Ma, Yihang Liu, Zhike Qiu, Jiayi Ji, Xiaoshuai Sun• 2026

Related benchmarks

TaskDatasetResultRank
Object Hallucination EvaluationPOPE--
2019
Multimodal EvaluationMME
MME Score1.90e+3
19
Diagram UnderstandingAI2D
AI2D Accuracy55.34
19
Multimodal ReasoningMMBench
MMBench Accuracy64.18
19
Visual Question AnsweringGQA
GQA Score61.43
19
Science Question AnsweringSQA
SQA Score69.11
19
Visual Question AnsweringVizWiz
Accuracy (VizWiz)54.38
19
Showing 7 of 7 rows

Other info

Follow for update