Decision Correctness on HMDA (test)
92.2F1 ScoreGB
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GBDescription=Gradient Boosting, Prompt Type=Expert2025.12 | 92.2 | 88.1 | 91 | 93.5 | |
| GBDescription=Gradient Boosting, Prompt Type=Consumer2025.12 | 92.2 | 88.1 | 91 | 93.5 | |
| XGBoostPrompt Type=Expert2025.12 | 91.7 | 87.5 | 92.2 | 91.1 | |
| XGBoostPrompt Type=Consumer2025.12 | 91.7 | 87.5 | 92.2 | 91.1 | |
| GRPO-Step2Stage=Explanation Tone Tuning, Prompt Type=Expert2025.12 | 90.2 | 85.1 | 90 | 90.3 | |
| GRPO-Step1Stage=Decision Correctness Tuning, Prompt Type=Expert2025.12 | 89.7 | 84.5 | 90.3 | 89 | |
| GRPO-Step1Stage=Decision Correctness Tuning, Prompt Type=Consumer2025.12 | 89.3 | 84 | 89.8 | 88.9 | |
| GRPO-Step2Stage=Explanation Tone Tuning, Prompt Type=Consumer2025.12 | 89.3 | 82.5 | 83 | 96.7 | |
| NNDescription=Neural Network, Prompt Type=Expert2025.12 | 88.9 | 83.3 | 89.5 | 88.2 | |
| NNDescription=Neural Network, Prompt Type=Consumer2025.12 | 88.9 | 83.3 | 89.5 | 88.2 | |
| LRDescription=Logistic Regression, Prompt Type=Expert2025.12 | 85.8 | 75.2 | 75.6 | 99.3 | |
| LRDescription=Logistic Regression, Prompt Type=Consumer2025.12 | 85.8 | 75.2 | 75.6 | 99.3 | |
| SFTStage=Supervised Fine-Tuning, Prompt Type=Expert2025.12 | 82.5 | 75.7 | 90.8 | 75.6 | |
| SFTStage=Supervised Fine-Tuning, Prompt Type=Consumer2025.12 | 79.3 | 71.7 | 88.6 | 71.8 | |
| GPT-5Prompt Type=Consumer2025.12 | 77.1 | 67.9 | 84 | 71.3 | |
| GPT-5Prompt Type=Expert2025.12 | 73 | 63.9 | 84.4 | 64.2 | |
| Raw QwenPrompt Type=Consumer2025.12 | 72.6 | 63.2 | 83.1 | 64.5 | |
| Raw QwenPrompt Type=Expert2025.12 | 72.3 | 62.7 | 82.8 | 64.1 |