Journal tier classification on Economics cross-field (val)
69.5AccuracyQwen3-30B SFT (economics)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3-30B SFT (economics)N=200, Role=In-domain SFT2026.03 | 69.5 | 70.4 | |
| Qwen3-30B pooled (mgmt + econ)N=200, Role=Pooled SFT2026.03 | 69.5 | 70.5 | |
| GPT-4.1-nano SFT (economics)N=200, Role=In-domain SFT2026.03 | 68.5 | 69.4 | |
| GPT-4.1-nano pooled (mgmt + econ)N=200, Role=Pooled SFT2026.03 | 67 | 67.4 | |
| Qwen3-4B SFT (economics)N=200, Role=In-domain SFT2026.03 | 64 | 64.8 | |
| Mgmt SFT GPT-4.1 (cross-field)N=200, Role=Cross-field transfer2026.03 | 43.5 | 38.9 | |
| GPT-4.1 base (cross-field)N=200, Role=Base comparator2026.03 | 29.5 | — | |
| Qwen3-30B baseN=200, Role=Base control2026.03 | 25.5 | 16.9 | |
| GPT-4.1-nano baseN=200, Role=Base control2026.03 | 25 | 15.6 | |
| Qwen3-4B baseN=200, Role=Base control2026.03 | 25 | 10 |