Adherence evaluation on Medical Prompts (150 held-out prompts)
2.85Mean AdherenceCAFT-time vsvd
Evaluation results
| Method | Links | |
|---|---|---|
| CAFT-time vsvdEpochs=1 epoch2026.05 | 2.85 | |
| NaiveEpochs=1 epoch2026.05 | 2.28 | |
| GRASPEpochs=1 epoch2026.05 | 2.28 | |
| Inference-ablate vsvdInference-time ablation=true2026.05 | 1.18 |