Data Extraction on Finance D2
47.6Match Ratio (Mean)SFT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SFTBackbone=Qwen2.5-32B2025.05 | 47.6 | 76.5 | 40 | 68.6 | 76.8 | 75.5 | |
| GRPOBackbone=Qwen2.5-7B2025.05 | 43.5 | 74.9 | 35.6 | 68.8 | 76.2 | 71.5 | |
| GRPOBackbone=Llama3.1-8B2025.05 | 42.6 | 77.9 | 35.7 | 72.9 | 71.9 | 67.5 | |
| SFTBackbone=Qwen2.5-7B2025.05 | 40.9 | 71.6 | 32.9 | 64.4 | 74.7 | 70.5 | |
| SFTBackbone=Llama3.1-8B2025.05 | 37.6 | 67.3 | 30.5 | 61.1 | 70.4 | 68 | |
| RawBackbone=Qwen2.5-32B2025.05 | 23.7 | 38.2 | 10.8 | 24.1 | 72.2 | 63 | |
| GRPOBackbone=Llama3.2-3B2025.05 | 20.6 | 38.5 | 8.3 | 27.4 | 67 | 67.5 | |
| SFTBackbone=Llama3.2-3B2025.05 | 20.3 | 38.4 | 8.8 | 28 | 72.6 | 72 | |
| RawBackbone=Llama3.1-8B2025.05 | 19.5 | 28.5 | 7.7 | 16.9 | 66.9 | 50.5 | |
| RawBackbone=Qwen2.5-7B2025.05 | 18.6 | 31.6 | 6.8 | 19.5 | 66.1 | 57 | |
| RawBackbone=Llama3.2-3B2025.05 | 8.9 | 19.4 | 4 | 11.8 | 66.7 | 50 |