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

LearnPruner: Rethinking Attention-based Token Pruning in Vision Language Models

About

Vision-Language Models (VLMs) have recently demonstrated remarkable capabilities in visual understanding and reasoning, but they also impose significant computational burdens due to long visual sequence inputs. Recent works address this issue by pruning unimportant visual tokens, achieving substantial computational reduction while maintaining model performance. The core of token pruning lies in determining token importance, with current approaches primarily relying on attention scores from vision encoders or Large Language Models (LLMs). In this paper, we analyze the effectiveness of attention mechanisms in both vision encoders and LLMs. We find that vision encoders suffer from attention sink, leading to poor focus on informative foreground regions, while in LLMs, although prior studies have identified attention bias toward token positions, text-to-vision attention demonstrates resistance to this bias and enables effective pruning guidance in middle layers. Based on these observations, we propose LearnPruner, a two-stage token pruning framework that first removes redundant vision tokens via a learnable pruning module after the vision encoder, then retains only task-relevant tokens in the LLM's middle layer. Experimental results show that our LearnPruner can preserve approximately 95% of the original performance while using only 5.5% of vision tokens, and achieve 3.2$\times$ inference acceleration, demonstrating a superior accuracy-efficiency trade-off.

Rinyoichi Takezoe, Yaqian Li, Zihao Bo, Anzhou Hou, Mo Guang, Kaiwen Long• 2026

Related benchmarks

TaskDatasetResultRank
Object Hallucination EvaluationPOPE--
2019
Visual Question AnsweringVQA v2--
1429
Science Question AnsweringScienceQA--
791
Multimodal EvaluationMME--
727
Multimodal BenchmarkingMMBench CN
Score56.8
151
Text-based Visual Question AnsweringTextVQA VQAT
Accuracy57.3
61
Multimodal BenchmarkingMMBench
MMBench Score63.8
60
Multimodal EvaluationLLaVA Evaluation Suite 7B v1.5 (test)
GQA60.3
34
Visual Question AnsweringGQA
GQA Score60.3
26
Video Question AnsweringVideoQA Average
Score2.51
11
Showing 10 of 14 rows

Other info

Follow for update