Driving explanation classification on BDD-X Human 1
90.37AccuracyRACE
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| RACEPrompting Strategy=CoT-SC2026.02 | 90.37 | 0.85 | 0.89 | 0.83 | — | |
| GPT-4oPrompting Strategy=CoT2026.02 | 82.44 | 0.77 | 0.8 | 0.72 | — | |
| GPT-4o-miniPrompting Strategy=CoT2026.02 | 80.18 | 0.65 | 0.77 | 0.76 | — | |
| GPT-3.5-TurboPrompting Strategy=CoT2026.02 | 77.44 | 0.62 | 0.74 | 0.74 | — | |
| RACEPrompting Strategy=CoT-SC2026.02 | 60.22 | 0.47 | 0.57 | 0.83 | — | |
| GPT-4oPrompting Strategy=CoT2026.02 | 54.55 | 0.43 | 0.51 | 0.72 | — | |
| GPT-3.5-TurboPrompting Strategy=CoT2026.02 | 50.56 | 0.36 | 0.46 | 0.74 | — | |
| GPT-4o-miniPrompting Strategy=CoT2026.02 | 48.7 | 0.33 | 0.45 | 0.76 | — |