Benchmark RadarAI BENCHMARK PROFILE
VLMsAreBiased
VLMsAreBiased evaluates whether vision-language models rely on visual evidence or fall back on language priors when answering.
- Released
- Unknown
- Readiness
- Paper only
- Primary field
- General AI
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.