Mathematical Reasoning on AIME 2024 (mean@30, pass@30)
35.1Mean @30SFT & RL*
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SFT & RL*Backbone=Youtu-LLM-2B-Base, Training Stage=RL2026.06 | 35.1 | 56.7 | |
| FE-SFT & FC-RLBackbone=Youtu-LLM-2B-Base, Training Stage=RL, Framework=WM-AMT2026.06 | 34.6 | 66.7 | |
| SFT & RLBackbone=Youtu-LLM-2B-Base, Training Stage=RL, Framework=WM-AMT2026.06 | 34.2 | 63.3 | |
| FE-SFT & RLBackbone=Youtu-LLM-2B-Base, Training Stage=RL, Framework=WM-AMT2026.06 | 33.6 | 60 | |
| IWM & RLBackbone=Youtu-LLM-2B-Base, Training Stage=RL2026.06 | 33.2 | 66.7 | |
| State-SFT & RL‡Backbone=Youtu-LLM-2B-Base, Training Stage=RL, Framework=WM-AMT2026.06 | 33.2 | 63.3 | |
| State-SFT & RL†Backbone=Youtu-LLM-2B-Base, Training Stage=RL2026.06 | 32.2 | 60 | |
| FE-SFT & RL*Backbone=Youtu-LLM-2B-Base, Training Stage=RL2026.06 | 31.9 | 60 |