MEGA Hub

DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models

Authors

Do you know Chen Zhong?You can claim authorship or link another user.Do you know Xiao An?You can claim authorship or link another user.Do you know Zijie Wang?You can claim authorship or link another user.Do you know Jiepan Li?You can claim authorship or link another user.Do you know Guangyi Yang?You can claim authorship or link another user.Do you know Wei He?You can claim authorship or link another user.

Abstract

Visual inputs in vision-language models (VLMs) are often encoded into substantially longer token sequences than text, making visual tokens a major bottleneck for efficient inference. Abundant recent methods address this bottleneck by scoring token importance and pruning low-scoring tokens in a single pass. However, one-shot scoring is insufficient because a token's prompt-relevant usefulness depends on the evidence already retained. Motivated by this insight, we introduce DIVE (Dynamic Iterative Visual Evidence Construction), a training-free framework that recasts visual-token pruning as dynamic evidence construction. DIVE repeatedly selects the remaining token with the highest residual-conditioned score, updates the visual and prompt residuals to discount the evidence already explained, and re-evaluates the remaining tokens. This select-update-re-evaluate process builds a retained set of complementary, prompt-relevant evidence. Experiments across eight image-understanding benchmarks show that DIVE consistently preserves performance across token budgets. With an 88.9% reduction in visual tokens, DIVE retains 98.2% of the uncompressed model's average performance. Code is available at https://github.com/Zhong-Chenchen/DIVE.git.

Community

00