CVSep 29, 2025

VideoAnchor: Reinforcing Subspace-Structured Visual Cues for Coherent Visual-Spatial Reasoning

arXiv:2509.25151v11 citationsh-index: 5Has Code
Originality Highly original
AI Analysis

This addresses the problem of inconsistent visual-spatial reasoning in MLLMs for researchers and practitioners, representing an incremental improvement through a novel method for a known bottleneck.

The paper tackles the limitation of Multimodal Large Language Models in visual-spatial reasoning by proposing VideoAnchor, a plug-and-play module that reinforces visual cues across frames, resulting in performance gains such as 3.2% and 4.6% improvements on benchmarks like VSI-Bench and Video-MME.

Multimodal Large Language Models (MLLMs) have achieved impressive progress in vision-language alignment, yet they remain limited in visual-spatial reasoning. We first identify that this limitation arises from the attention mechanism: visual tokens are overshadowed by language tokens, preventing the model from consistently recognizing the same visual cues across frames. To address this challenge, we draw a novel connection between the self-expressiveness property in sparse subspace clustering and the attention mechanism in Transformers. Building on this insight, we propose VideoAnchor, a plug-and-play module that leverages subspace affinities to reinforce visual cues across frames without retraining, effectively anchoring attention to shared visual structures. Extensive experiments across benchmarks and backbone models show consistent performance gains -- $e.g.$, 3.2% and 4.6% improvements on VSI-Bench and Video-MME (spatial-related tasks) with InternVL2-8B and Qwen2.5VL-72B -- while qualitative analyses demonstrate more coherent subspace partitions and stronger visual grounding. Our codes will be made public available at https://github.com/feufhd/VideoAnchor.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes