Mathematical Reasoning on GSM8K (Win rate)
62.97Win RateTOSS
Evaluation Results
| Method | Links | |
|---|---|---|
| TOSSBackbone=Llama-3-8B-Instruct2026.03 | 62.97 | |
| SEALBackbone=Llama-3-8B-Instruct2026.03 | 60.62 | |
| Standard SFTBackbone=Llama-3-8B-Instruct2026.03 | 58.59 |
Loading the SOTA2 catalog…