EM Recognition (PR) on PReD-Bench
72MCQA ScorePReD
Evaluation Results
| Method | Links | |
|---|---|---|
| PReD2026.03 | 72 | |
| gemini-2.5 pro2026.03 | 24 | |
| claude sonnet-42026.03 | 21.6 | |
| seed-1.6 vision2026.03 | 21.4 | |
| GPT-52026.03 | 19.2 | |
| qwen2.5-vl-7bInstruction-tuned=true2026.03 | 14 | |
| qwen3-vl-235b-a22bInstruction-tuned=true2026.03 | 8.6 | |
| qwen3-vl-8bInstruction-tuned=true2026.03 | 2.8 |