multimodal benchmark
VLMsAreBiased
VLMsAreBiased evaluates whether vision-language models rely on visual evidence or fall back on language priors when answering.
Updated Aug 11, 2026
Models2
Model coverage2
MetricScore
EvidenceB
VLMsAreBiased Ranking
Higher score ranks better on this benchmark.
| 1 | BY | 83.6% | 100.0% | 2 | C | |
| 2 | BY | 68.3% | 0.0% | 2 | C |
VLMsAreBiased Score Distribution
A closer view of the leading scores on this benchmark.
VLMsAreBiased Highlights
The leading models and scores on this benchmark.
What is VLMsAreBiased?
What VLMsAreBiased measures and how its scores work.
VLMsAreBiased evaluates whether vision-language models rely on visual evidence or fall back on language priors when answering.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- VLMsAreBiased
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- vlmsarebiased|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about VLMsAreBiased.
Which model scores highest on VLMsAreBiased?
Seed 2.1 Pro is currently ranked first with 83.6%.
What does VLMsAreBiased measure?
VLMsAreBiased evaluates whether vision-language models rely on visual evidence or fall back on language priors when answering.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
2 model results are currently shown.
Does this benchmark affect the overall score?
No. This benchmark is shown for reference but does not contribute to the overall score.