EM Decision-Making (AJSD) on PReD-Bench
0.253BLEU-4 ScorePReD
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PReDEvaluation Protocol=OpenQA#2026.03 | 0.253 | 0.559 | 0.515 | 0.617 | |
| qwen2.5-vl-7bInstruction-tuned=true, Evaluation Protocol=OpenQA#2026.03 | 0 | 0.079 | 0.115 | 0.128 | |
| qwen3-vl-8bInstruction-tuned=true, Evaluation Protocol=OpenQA#2026.03 | 0 | 0.094 | 0.121 | 0.228 | |
| qwen3-vl-235b-a22bInstruction-tuned=true, Evaluation Protocol=OpenQA#2026.03 | 0 | 0.09 | 0.07 | 0.151 | |
| seed-1.6 visionEvaluation Protocol=OpenQA#2026.03 | 0 | 0.095 | 0.039 | 0.169 | |
| claude sonnet-4Evaluation Protocol=OpenQA#2026.03 | 0 | 0.103 | 0.078 | 0.154 | |
| GPT-5Evaluation Protocol=OpenQA#2026.03 | 0 | 0.017 | 0.028 | 0.055 | |
| gemini-2.5 proEvaluation Protocol=OpenQA#2026.03 | 0 | 0.081 | 0.036 | 0.129 |