Downstream Policy Evaluation on OLMo3 Adapt
96GSM8KSequential
Evaluation Results
| Method | Links | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SequentialTraining=RC2026.05 | 96 | 93.4 | 80.2 | 67.7 | 64.7 | 85.8 | 36.8 | 32.3 | 57.1 | 83.2 | 87.4 | 35.1 | 68.3 | |
| SequentialTraining=IQ2026.05 | 95.9 | 92.7 | 76.5 | 66.7 | 64.6 | 85 | 38.1 | 32 | 54 | 83 | 86.9 | 41.1 | 68 | |
| EVOLM2026.05 | 95.8 | 94.5 | 86.2 | 68.5 | 67.6 | 85.3 | 37.7 | 30.5 | 54.2 | 82.3 | 86.4 | 42.2 | 69.3 | |
| GPT-4.1Prompted=true2026.05 | 95.7 | 94.2 | 58.7 | 65.1 | 71.7 | 84.8 | 34.4 | 30.7 | 54.9 | 81.9 | 82.9 | 45.5 | 66.7 | |
| RAR2026.05 | 95.7 | 93.8 | 77.4 | 68.6 | 62.5 | 84.8 | 36 | 32.2 | 52.5 | 84.1 | 87.3 | 34.7 | 67.5 | |
| Scalar RMBackbone=Skywork-RM-V22026.05 | 95.6 | 94.3 | 47.9 | 66.2 | 54.9 | 82 | 37.3 | 25.2 | 46.2 | 56.4 | 81.6 | 28.3 | 59.7 | |
| RUBRIC-ARM2026.05 | 95.6 | 93.3 | 78.4 | 62.3 | 64.4 | 83.8 | 35.7 | 32.8 | 56.9 | 83.1 | 87.2 | 35.9 | 67.5 | |
| Qwen3-8BPrompted=true2026.05 | 95.5 | 94.6 | 73.7 | 58.4 | 66.8 | 83.4 | 35.3 | 31.7 | 56.9 | 84.3 | 83.9 | 45.9 | 67.5 | |
| RRD2026.05 | 95.3 | 92.2 | 75.2 | 59.2 | 69.4 | 84.8 | 36.8 | 32.6 | 56 | 82.6 | 88.9 | 38.5 | 67.6 | |
| RLCER2026.05 | 94 | 94.7 | 80.5 | 67.9 | 68.6 | 84.8 | 39.2 | 30.4 | 49.6 | 69.1 | 84.3 | 37.2 | 66.7 |