multimodal benchmark
VQAv2
VQAv2 is a balanced Visual Question Answering dataset that addresses language bias by providing complementary images for each question, forcing models to rely on visual understanding rather than language priors. It contains approximately twice the number of image-question pairs compared to the original VQA dataset.
Updated Aug 7, 2026
Published models3
Registry coverage3
MetricScore
EvidenceB
VQAv2 leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | MA | 80.9% | 100.0% | 3 | C | |
| 2 | MA | 78.6% | 50.0% | 3 | C | |
| 3 | ME | 78.1% | 0.0% | 3 | C |
Score distribution
Top published rows on the benchmark's original scale.
VQAv2 highlights
The top published results on this benchmark's own scale.
What is VQAv2?
Definition and scoring fields from the benchmark registry.
VQAv2 is a balanced Visual Question Answering dataset that addresses language bias by providing complementary images for each question, forcing models to rely on visual understanding rather than language priors. It contains approximately twice the number of image-question pairs compared to the original VQA dataset.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- VQAv2
- Modality
- multimodal
- Primary category
- multimodal
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- vqav2|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about VQAv2.
Which model scores highest on VQAv2?
Pixtral Large is currently ranked first with 80.9%.
What does VQAv2 measure?
VQAv2 is a balanced Visual Question Answering dataset that addresses language bias by providing complementary images for each question, forcing models to rely on visual understanding rather than language priors. It contains approximately twice the number of image-question pairs compared to the original VQA dataset.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
3 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.