Rationale Quality Evaluation on UX-car
3.91UI GroundingPERCEPTUI
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| PERCEPTUI2026.06 | 3.91 | 3.74 | 3.88 | 3.94 | |
| PERCEPTUIvariant=w/o RPE2026.06 | 3.68 | 3.56 | 3.61 | 3.71 | |
| GPT-52026.06 | 3.61 | 3.32 | 3.47 | 3.56 | |
| GPT-4o2026.06 | 3.42 | 3.05 | 3.11 | 3.28 | |
| Claude Opus 4.62026.06 | 3.34 | 3.01 | 3.13 | 3.29 | |
| Gemini-2.5-pro2026.06 | 3.26 | 2.88 | 3.02 | 3.17 | |
| SFT2026.06 | 2.61 | 2.38 | 2.29 | 2.74 | |
| PERCEPTUIvariant=w/o CR2026.06 | 1.7 | 1.55 | 1.61 | 1.89 |