Physical Code Generation on PhysCodeBench User Study
4.5Physical AccuracySMRF + SFT + DPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SMRF + SFT + DPO2026.04 | 4.5 | 4.2 | 4.6 | |
| DRDQ-32B + SFT + DPO2026.04 | 3.4 | 3.5 | 3.6 | |
| Claude-3.5-Sonnet2026.04 | 3.2 | 3.7 | 3.4 |
| Method | Links | |||
|---|---|---|---|---|
| SMRF + SFT + DPO2026.04 | 4.5 | 4.2 | 4.6 | |
| DRDQ-32B + SFT + DPO2026.04 | 3.4 | 3.5 | 3.6 | |
| Claude-3.5-Sonnet2026.04 | 3.2 | 3.7 | 3.4 |