Vision Reasoning on TRACER-Bench
100LT ScoreHuman
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Human2026.05 | 100 | 97.6 | 95.2 | 99 | |
| GPT-5.22026.05 | 58.4 | 68.2 | 52.2 | 72.7 | |
| Gemini 3.1 Pro2026.05 | 53.5 | 64.5 | 44.1 | 67.8 | |
| Qwen3-VLParam=8B, Instruct=true2026.05 | 35.9 | 55.1 | 15.1 | 52.7 | |
| CAVE-7BParam=7B, Instruct=false2026.05 | 31.4 | 58.4 | 20.8 | 59.2 | |
| MiniCPM-V 4.5Param=8B, Instruct=false2026.05 | 30.2 | 52.7 | 16.7 | 55.1 | |
| CAVE Belief-OnlyParam=7B, Instruct=false2026.05 | 28.8 | 52.8 | 16.8 | 55.2 | |
| CAVE-3BParam=3B, Instruct=false2026.05 | 27.2 | 56 | 18.4 | 57.6 | |
| InternVL3.5Param=8B, Instruct=true2026.05 | 26.1 | 41.2 | 23.7 | 58.4 | |
| DeepEyesParam=7B, Instruct=false2026.05 | 23.7 | 49.4 | 15.9 | 54.3 | |
| Qwen2.5-VLParam=7B, Instruct=true2026.05 | 22.9 | 53.5 | 15.9 | 43.7 | |
| R1-OneVisionParam=7B, Instruct=false2026.05 | 22 | 43.7 | 7.8 | 51.8 | |
| VLAA-ThinkerParam=7B, Instruct=false2026.05 | 22 | 43.7 | 14.3 | 47.8 | |
| ThinkLite-VLParam=7B, Instruct=false2026.05 | 22 | 45.3 | 17.6 | 50.2 | |
| Random2026.05 | 16.2 | 50 | 12.1 | 50 |