PapersSelected73°

VIP-Router for Vision Token Pruning

Beyond One-Size-Fits-All: Sample-Adaptive Strategy Routing for Vision Token Pruning in MLLMs

Why we picked it

VIP-Router reduces inference costs for MLLMs while maintaining accuracy, offering a plug-and-play solution that works across different model backbones.

VIP-Router is a lightweight vision pruning router that adaptively selects the best pruning strategy for each input in multimodal large language models.

Body

Evaluated on VTC-Bench Group A, it achieves a 26.9% relative improvement in average accuracy compared to fixed strategies.

The model introduces only 0.017% trainable parameters and operates without modifying underlying algorithms or weights.

Original · arXiv cs.AI

Beyond One-Size-Fits-All: Sample-Adaptive Strategy Routing for Vision Token Pruning in MLLMs

Multimodal large language models (MLLMs) process hundreds or thousands of visual tokens per image, incurring prohibitive inference costs. While existing vision token pruning methods mitigate this overhead, they implicitly assume that a single fixed pruning strategy can be applied uniformly across all inputs. Our analysis further reveals that ranking pruning methods by average benchmark accuracy conceals substantial sample-wise complementarity: although the average-best strategy excels overall, alternative strategies prove superior on a significant fraction of individual samples. To harness this diversity, we propose VIP-Router, a lightweight VIsion Pruning Router that adaptively selects the pruning strategy predicted to be best suited to each input at a specified pruning level. Conditioned on low-cost visual and textual features, VIP-Router identifies the most suitable candidate strategy while retaining full-token inference as an option when pruning is predicted to be unfavorable. Evaluated on a curated suite of pruning-sensitive visual perception benchmarks, VTC-Bench Group A, VIP-Router consistently outperforms the best fixed strategy baseline across all reduction ratios, achieving a 26.9% relative improvement in average accuracy, and a 22.0% relative increase in average utility after accounting for realized token cost. Crucially, VIP-Router operates in a plug-and-play manner without modifying underlying pruning algorithms or model weights, introducing trainable parameters equivalent to merely 0.017\% of the backbone. Furthermore, VIP-Router proves effective across various MLLM backbones and yields consistent gains on unseen benchmarks, highlighting the potential of sample adaptive routing for visual token pruning.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Papers" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include MLLMs, VIP-Router, multimodal.

Source attribution: curated from arXiv cs.AI (@Haiji Liang, Pengfei Zhou, Zhenglin Wan, Wei Wang, Yang You, Wangbo Zhao).