Mathematical Reasoning on AMC 23 (Avg@16, #Tokens)
93Average Accuracy @16GR3
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GR3Model size=7B2026.03 | 93 | 3,090 | |
| DLER–R1–7BModel size=7B2026.03 | 91.4 | 2,255 | |
| GRPOModel size=7B2026.03 | 90.3 | 7,256 | |
| DeepSeek–R1–Distill–7BModel size=7B2026.03 | 89.8 | 6,385 | |
| Laser–DE–L4096–7BModel size=7B2026.03 | 88.1 | 2,427 | |
| AdaptThink–7BModel size=7B2026.03 | 88.1 | 4,280 | |
| LCR1–7BModel size=7B2026.03 | 85.8 | 2,963 |