Cumulative temporal regret on CAP-GSM8K
21.3RLVR R_temp CERLVR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RLVRModel=Claude-Sonnet-4.5, E=100, seeds=32026.06 | 21.3 | — | — | |
| RLVRModel=GPT-4o, E=100, seeds=32026.06 | 58 | — | — | |
| RLVRModel=GPT-3.5-Turbo, E=100, seeds=32026.06 | 68 | — | — | |
| RLVRModel=Llama-3.3-70b, E=500, seeds=102026.06 | 248.8 | — | — | |
| TriviumModel=Llama-3.3-70b, E=500, seeds=102026.06 | — | 2.4 | 103 | |
| TriviumModel=GPT-4o, E=100, seeds=32026.06 | — | 3 | 19 | |
| TriviumModel=Claude-Sonnet-4.5, E=100, seeds=32026.06 | — | 2.3 | 9 | |
| TriviumModel=GPT-3.5-Turbo, E=100, seeds=32026.06 | — | 3.7 | 18 |