Beyond Intermediate States: Explaining Visual Redundancy through Language
Reviewed by Pithpith:MTOQ2BKGopen to challenge →
read the original abstract
Multi-modal Large Langue Models (MLLMs) often process thousands of visual tokens, which consume a significant portion of the context window and impose a substantial computational burden. Prior work has empirically explored visual token pruning methods based on MLLMs' intermediate states (e.g., attention scores). However, they have limitations in precisely defining visual redundancy due to their inability to capture the influence of visual tokens on MLLMs' visual understanding (i.e., the predicted probabilities for textual token candidates). To address this issue, we manipulate the visual input and investigate variations in the textual output from both token-centric and context-centric perspectives, achieving intuitive and comprehensive analysis. Experimental results reveal that visual tokens with low ViT-[cls] association and low text-to-image attention scores can contain recognizable information and significantly contribute to images' overall information. To develop a more reliable method for identifying and pruning redundant visual tokens, we integrate these two perspectives and introduce a context-independent condition to identify redundant prototypes from training images, which probes the redundancy of each visual token during inference. Extensive experiments on single-image, multi-image and video comprehension tasks demonstrate the effectiveness of our method, notably achieving 90% to 110% of the performance while pruning 80% to 90% of visual tokens.
This paper has not been read by Pith yet.
Forward citations
Cited by 2 Pith papers
-
A Cross-Modal Prompt Injection Attack against Large Vision-Language Models with Image-Only Perturbation
CrossMPI steers both visual and textual interpretations in LVLMs through image-only perturbations by optimizing in hidden-state space at selected middle layers with distance-based budget allocation.
-
Visual-Redundancy-Controlled Parallel Decoding for Diffusion-Based Multimodal Large Language Models
VRCD prioritizes visually complementary positions during parallel decoding in dMLLMs by measuring attention overlap with the new Visual Redundancy Index, yielding accuracy gains over confidence-based baselines on M^3C...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.