Jigsaw on BLINK Jigsaw
56.8AccuracyBinary verification (Certainty)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Binary verification (Certainty)Model=GPT-4o, Inference Mode=True/False procedure, Policy=Certainty (with uncertain → False policy)2025.11 | 56.8 | 55.5 | |
| Binary verification (Normal)Model=GPT-4o, Inference Mode=True/False procedure, Policy=Normal (without uncertain → False policy)2025.11 | 52.8 | 51.4 | |
| MCQ (Single-shot)Model=GPT-4o, Inference Mode=Single-shot2025.11 | 51.6 | 49.6 | |
| MCQ (MV)Model=GPT-4o, Inference Mode=3× MV (majority vote with three repeats)2025.11 | 45.2 | 43.9 |