Benchmark
BLINK - Visual_Correspondence
Vision benchmarks measure how well models understand images. Scores are percentages.
All vision benchmarksScores from EvalAI data.
- Models measured
- 4
- Scores recorded
- 4
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 29.7%
- Mean
- 31.4%
- Median
- 5.3%
- Std dev
- 22.7%
- 10th percentile
- 37.2%
- 90th percentile
- 22.7%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Gemini ProVisual_Correspondence· EvalAI data | 37.2% |
| 2 | Claude 3 OpusVisual_Correspondence· EvalAI data | 31.4% |
| 3 | LLaVA 1.6 34BVisual_Correspondence· EvalAI data | 27.3% |
| 4 | Qwen-VL MaxVisual_Correspondence· EvalAI data | 22.7% |