Mathematical Reasoning on GSM8K DE
63Compositional AccuracyHATified-SFT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| HATified-SFTMode=SFT, Shots=8-shot2026.03 | 63 | — | |
| T-Freeshots=8-shot2026.03 | 58 | 4.42 | |
| Tülushots=8-shot2026.03 | 56.8 | 2.94 | |
| HATifiedshots=8-shot2026.03 | 53.4 | 4.46 | |
| Llamashots=8-shot2026.03 | 45.9 | 2.94 | |
| Llama-InstructMode=SFT, Shots=8-shot2026.03 | 13.9 | — |