Mathematical Reasoning on AMC23 (Accuracy and Length)
2,246Solution LengthShorterBetter
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ShorterBetterBackbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=3002026.03 | 2,246 | 66.9 | |
| ShorterBetterBackbone=DeepSeek-R1-Distill-Qwen-7B, #RL Steps=2002026.03 | 2,719 | 86.3 | |
| LASER-DE-4096Backbone=DeepSeek-R1-Distill-Qwen-7B, #RL Steps=10002026.03 | 2,946 | 90 | |
| Truncated Think 4kBackbone=DeepSeek-R1-Distill-Qwen-7B, #RL Steps=N/A2026.03 | 3,511 | 45.6 | |
| Truncated Think 4kBackbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=N/A2026.03 | 3,627 | 42.5 | |
| SmartThinkerBackbone=DeepSeek-R1-Distill-Qwen-7B, #RL Steps=752026.03 | 4,118 | 90 | |
| ThinkPrune-4kBackbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=N/A2026.03 | 4,201 | 76.3 | |
| Truncated Think 4kBackbone=Qwen3-4B-Thinking-2507, #RL Steps=N/A2026.03 | 4,319 | 12 | |
| LASER-DE-4096Backbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=10002026.03 | 4,330 | 71.9 | |
| SmartThinkerBackbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=1502026.03 | 4,421 | 76.3 | |
| Truncated Think 8kBackbone=DeepSeek-R1-Distill-Qwen-7B, #RL Steps=N/A2026.03 | 4,657 | 53.1 | |
| Truncated Think 8kBackbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=N/A2026.03 | 5,471 | 50.6 | |
| SmartThinkerBackbone=Qwen3-4B-Thinking-2507, #RL Steps=502026.03 | 5,992 | 98.8 | |
| Base ModelBackbone=DeepSeek-R1-Distill-Qwen-7B, #RL Steps=N/A2026.03 | 6,634 | 91.9 | |
| Truncated Think 8kBackbone=Qwen3-4B-Thinking-2507, #RL Steps=N/A2026.03 | 7,067 | 15 | |
| Base ModelBackbone=DeepSeek-R1-Distill-Qwen-1.5B, #RL Steps=N/A2026.03 | 9,320 | 73.1 | |
| Base ModelBackbone=Qwen3-4B-Thinking-2507, #RL Steps=N/A2026.03 | 10,777 | 99.4 |