← Benchmarks
VLMsAreBiased
VLMsAreBiased evaluates whether vision-language models rely on visual evidence or fall back on language priors when answering.
id vlmsarebiased · max 1 · 4 models reported
| # | Model | Score |
|---|
No scores for this benchmark yet.