Reasoning on GSM8K (ERR)
10Error RateFLEx
Evaluation Results
| Method | Links | |
|---|---|---|
| FLExBackbone Model=Qwen-0.5B2026.01 | 10 | |
| FLExBackbone Model=Gemma-1B2026.01 | 70 | |
| FLExBackbone Model=Gemma-4B2026.01 | 417 | |
| FLExBackbone Model=Qwen-3B2026.01 | 563 | |
| FLExBackbone Model=Gemma-12B2026.01 | 625 | |
| FLExBackbone Model=Qwen-1.5B2026.01 | 625 | |
| FLExBackbone Model=Qwen-7B2026.01 | 775 | |
| FLExBackbone Model=Gemma-27B2026.01 | 2,208 | |
| FLExBackbone Model=Qwen-72B2026.01 | 2,963 | |
| FLExBackbone Model=Qwen-14B2026.01 | 3,784 | |
| FLExBackbone Model=Qwen-32B2026.01 | 5,385 |