Mathematical Expression Synthesis on GSM-Symbolic (test)
66Accuracy (%)SEVerA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SEVerAParameter tuning=with, base_model=Qwen3-8B2026.03 | 66 | 0 | 16.7 | |
| SEVerAParameter tuning=no, base_model=Qwen3-8B2026.03 | 53.2 | 0 | 18.8 | |
| CRANEbase_model=Qwen3-8B2026.03 | 44.7 | 2.1 | 12.4 | |
| LLM (Qwen3-8B)base_model=Qwen3-8B2026.03 | 38.3 | 10.6 | 10.9 |