Financial Language Modeling on Japanese LM Fin Harness
72.5Average Scoregpt-oss-20b + CPT + SFT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| gpt-oss-20b + CPT + SFTCoT=Yes2026.03 | 72.5 | 91.89 | 94.24 | 45.51 | 62.71 | 68.15 | |
| Qwen3-14B-Base + CPT + SFTCoT=Yes2026.03 | 71.78 | 91.62 | 91.45 | 48.59 | 60 | 67.27 | |
| Qwen3-14BCoT=Yes2026.03 | 71.04 | 91.96 | 93.26 | 49.37 | 53.37 | 67.22 | |
| Qwen3-14B-Base + SFTCoT=Yes2026.03 | 70.69 | 91.48 | 91.2 | 47.96 | 55 | 67.81 | |
| gpt-oss-20b + SFTCoT=Yes2026.03 | 69.56 | 91.87 | 90.87 | 43.65 | 52.63 | 68.8 | |
| gpt-oss-20bCoT=Yes2026.03 | 66.93 | 91.8 | 90.46 | 38.51 | 49.74 | 64.15 | |
| Qwen3-14BCoT=No2026.03 | 66.5 | 91.73 | 87.58 | 41.83 | 47.03 | 64.31 | |
| gpt-oss-20bCoT=No2026.03 | 61.17 | 91.09 | 81.33 | 33.45 | 41.82 | 58.17 |