Structured reward evaluation on AMC (test)
74.1AccuracyGAC + Token-φ
Evaluation Results
| Method | Links | |
|---|---|---|
| GAC + Token-φModel Size=14B, Seeds=32026.05 | 74.1 | |
| GAC w/o φModel Size=14B, Seeds=32026.05 | 73.2 | |
| HPTModel Size=14B, Seeds=32026.05 | 70.8 | |
| CHORDModel Size=14B, Seeds=32026.05 | 68.4 | |
| GAC + Token-φModel Size=7B, Seeds=32026.05 | 67.2 | |
| GAC w/o φModel Size=7B, Seeds=32026.05 | 65.8 | |
| HPTModel Size=7B, Seeds=32026.05 | 63.4 | |
| CHORDModel Size=7B, Seeds=32026.05 | 62.5 | |
| GAC + Token-φModel Size=1.5B, Seeds=32026.05 | 51.8 | |
| GAC w/o φModel Size=1.5B, Seeds=32026.05 | 51.4 | |
| HPTModel Size=1.5B, Seeds=32026.05 | 49.6 | |
| CHORDModel Size=1.5B, Seeds=32026.05 | 48.2 |