Human Behavioral Modeling on Human Behavioral Data Observed Features (held-out humans)
0.03Energy Distance (Sampling)Centaur
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CentaurParameters=70B2026.05 | 0.03 | 0.52 | 0.56 | 0.37 | |
| P-SFTEvaluation Setting=observed features2026.05 | 0.11 | 0.06 | 0.06 | 0.07 | |
| GPT-52026.05 | 0.29 | 5.84 | 2.28 | 2.81 | |
| A-SFTEvaluation Setting=observed features2026.05 | 0.55 | 0.53 | 0.97 | 0.68 | |
| A-SFTEvaluation Setting=cross-task2026.05 | 1.05 | 0.39 | 0.83 | 0.76 | |
| P-SFTEvaluation Setting=cross-task2026.05 | 1.28 | 0.44 | 0.59 | 0.77 | |
| Claude Sonnet2026.05 | 1.33 | 11.14 | 0.61 | 4.36 | |
| Base QwenParameters=1.5B2026.05 | 1.5 | 1.27 | 1.04 | 1.27 | |
| Gemini 2.5 Pro2026.05 | 2.06 | 1.3 | 1.42 | 1.59 |