Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2601.04946 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910280630403072 |
|---|---|
| author | Roy, Subhadeep Bhatia, Gagan Eger, Steffen |
| author_facet | Roy, Subhadeep Bhatia, Gagan Eger, Steffen |
| contents | Automatic metrics are widely used to evaluate text-to-image models, often replacing human judgment in benchmarking, model selection, and large-scale data filtering. Yet they may reward images that look plausible or prototypical rather than images that faithfully satisfy the prompt. We identify prototypicality bias as a systematic blindspot in multimodal evaluation: metrics can prefer a semantically incorrect but visually or socially prototypical image over a correct but less prototypical one. We introduce PROTOBIAS, a controlled diagnostic benchmark across Animals, Objects, and Demography, where semantically correct images are contrasted with plausible prototypical adversaries containing a single controlled semantic violation. Grounded in prototype theory and social-category prototypicality, PROTOBIAS is constructed with multiple prompt generators, image generators, and independent VLM filters, and validated through prompt-quality, human-annotation, and image-quality controls. Using PROTOBIAS, we show that widely used embedding, reward, VQA-based, and VLM-as-judge metrics frequently fail these contrasts, while human judgments remain more faithful to semantic correctness. We further introduce PROTOSCORE, a lightweight contrastively trained evaluator, as an initial mitigation baseline. PROTOBIAS provides a focused benchmark for measuring prototypicality-driven metric failures and developing more semantically faithful T2I evaluators. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_04946 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Prototypicality Bias Reveals Blindspots in Multimodal Evaluation Metrics Roy, Subhadeep Bhatia, Gagan Eger, Steffen Computer Vision and Pattern Recognition Artificial Intelligence Automatic metrics are widely used to evaluate text-to-image models, often replacing human judgment in benchmarking, model selection, and large-scale data filtering. Yet they may reward images that look plausible or prototypical rather than images that faithfully satisfy the prompt. We identify prototypicality bias as a systematic blindspot in multimodal evaluation: metrics can prefer a semantically incorrect but visually or socially prototypical image over a correct but less prototypical one. We introduce PROTOBIAS, a controlled diagnostic benchmark across Animals, Objects, and Demography, where semantically correct images are contrasted with plausible prototypical adversaries containing a single controlled semantic violation. Grounded in prototype theory and social-category prototypicality, PROTOBIAS is constructed with multiple prompt generators, image generators, and independent VLM filters, and validated through prompt-quality, human-annotation, and image-quality controls. Using PROTOBIAS, we show that widely used embedding, reward, VQA-based, and VLM-as-judge metrics frequently fail these contrasts, while human judgments remain more faithful to semantic correctness. We further introduce PROTOSCORE, a lightweight contrastively trained evaluator, as an initial mitigation baseline. PROTOBIAS provides a focused benchmark for measuring prototypicality-driven metric failures and developing more semantically faithful T2I evaluators. |
| title | Prototypicality Bias Reveals Blindspots in Multimodal Evaluation Metrics |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2601.04946 |