Almanac
paper

Gaze Heads: How VLMs Look at What They Describe

paperactiveprovisionalgaze-heads-how-vlms-look-at-what-they-describe-f3ab77a7·1 events·first seen 2d ago

Aliases: Gaze Heads: How VLMs Look at What They Describe

Co-occurring entities

More like this (12)

Recent events (1)

6arXiv · cs.CL·2d ago·source ↗

Gaze Heads: Attention heads in VLMs that track and control image region description

Researchers identify a small set of attention heads in vision-language model backbones, called 'gaze heads', whose attention patterns track the image region currently being described. Using comic strips as a controlled testbed, they show that intervening on the top-100 gaze heads (fewer than 9% of all heads) can steer the model to describe any chosen region at 83.1% accuracy, without retraining. The mechanism generalizes across model sizes from 2B to 32B parameters and to natural images (COCO), establishing a practical inference-time control lever for multimodal models via mechanistic analysis.