Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2605.11218 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910209895563264 |
|---|---|
| author | Shalankin, M. |
| author_facet | Shalankin, M. |
| contents | Embedded numeric anchors on images systematically bias Vision-Language Model quality judgments across six VLMs from five architectural families (ANOVA eta^2 = 0.18-0.77, all p < 0.001). Anchor effects are 2.5x larger than severe image quality degradation, confirming bias is not reducible to visual changes. Layer-wise probing reveals consistent dissociation: layers where anchor classification saturates (L12-L34) are suboptimal for quality prediction, with optimal layers deeper (R^2 = 0.69-0.91). Fusion analysis identifies architecture-dependent integration -- instant fusion at L1-L2 in two models versus partial or no fusion in three others. These results establish a causal account of visual anchoring bias, linking behavioral susceptibility to representation dynamics. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_11218 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Don't Look at the Numbers: Visual Anchoring Bias and Layer-wise Representation in VLMs Shalankin, M. Artificial Intelligence Embedded numeric anchors on images systematically bias Vision-Language Model quality judgments across six VLMs from five architectural families (ANOVA eta^2 = 0.18-0.77, all p < 0.001). Anchor effects are 2.5x larger than severe image quality degradation, confirming bias is not reducible to visual changes. Layer-wise probing reveals consistent dissociation: layers where anchor classification saturates (L12-L34) are suboptimal for quality prediction, with optimal layers deeper (R^2 = 0.69-0.91). Fusion analysis identifies architecture-dependent integration -- instant fusion at L1-L2 in two models versus partial or no fusion in three others. These results establish a causal account of visual anchoring bias, linking behavioral susceptibility to representation dynamics. |
| title | Don't Look at the Numbers: Visual Anchoring Bias and Layer-wise Representation in VLMs |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2605.11218 |