What Do Visual-Head Discovery Scores Actually Find? Attention Source, Necessity, and Random-Control Pitfalls in Vision-Language Models

Key points

• On Qwen3-VL-8B-Instruct we vary the query type, the discovery corpus and the token positions whose attention is read (the attention source) in a full 2 x 3 x 2 design, and test all 12 head-sets for steerability (CERS: Contrastive Evidence-Region Steering score) and necessity (NCS: Necessity Con- trast Score) against layer-matched random head-sets.1
• (2) Necessity follows the attention source: head-sets discovered from query-span attention are more nec- essary than last-token ones (difference 0.083, p=6 x 10^-10), and the two re- ferring + query-span sets lie 17-19 standard deviations above 250 layer-matched random sets (Holm-adjusted p=0.048).1
• The referring + query-span effect re- plicates on several recent VLMs. (3) Steerability is not specific to the discovered heads: no discovered head-set steers better than its layer-matched random sets (rank p >= 0.14), and random sets drawn from the same or from any layer span the discovered sets' scores.1
• We turn these findings into a protocol for reporting visual-head discoveries.1

Sources (1)

• [1]What Do Visual-Head Discovery Scores Actually Find? Attention Source, Necessity, and Random-Control Pitfalls in Vision-Language Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 07:06 AM

  Attention heads that attend to image regions relevant to a question are increasingly used to steer, ground and interpret vision-language models (VLMs) and to reduce their hallucinations.

  On Qwen3-VL-8B-Instruct we vary the query type, the discovery corpus and the token positions whose attention is read (the attention source) in a full 2 x 3 x 2 design, and test all 12 head-sets for steerability (CERS: Contrastive Evidence-Region Steering score) and necessity (NCS: Necessity Con- trast Score) against layer-matched random head-sets.

Extractive summary: sentences quoted from the sources.