Legal Document Generation on Legal Document Drafting Dataset (test)
0.15RL ScoreWrapper (Over LLaMA-3.1-8B)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Wrapper (Over LLaMA-3.1-8B)Base Model=LLaMA-3.1-8B, Prompting Strategy=Wrapper2025.04 | 0.15 | 0.04 | 0.18 | 0.79 | 0.19 | 5.15 | 3.3 | 2.2 | |
| Wrapper (Over Gemma-3-12B)Base Model=Gemma-3-12B, Prompting Strategy=Wrapper2025.04 | 0.15 | 0.06 | 0.24 | 0.8 | 0.17 | 6.56 | 8.82 | 7.82 | |
| GPT-4oBase Model=GPT-4o2025.04 | 0.14 | 0.03 | 0.12 | 0.81 | 0.24 | 6.68 | 8.8 | 5.4 | |
| Gemma-3-12B-It SFTBase Model=Gemma-3-12B, Prompting Strategy=Instruct, Fine-tuning Strategy=SFT2025.04 | 0.11 | 0.01 | 0.1 | 0.78 | 0.04 | 1.37 | 1 | 1 | |
| Qwen3-14BBase Model=Qwen3-14B2025.04 | 0.09 | 0 | 0.07 | 0.73 | 0.01 | 3.56 | 1 | 1 | |
| LLaMA-3.1-8B-InstructBase Model=LLaMA-3.1-8B, Prompting Strategy=Instruct2025.04 | 0.09 | 0.01 | 0.11 | 0.78 | 0.04 | 1.57 | 1 | 1.1 | |
| Gemma-3-12B-ItBase Model=Gemma-3-12B, Prompting Strategy=Instruct2025.04 | 0.09 | 0.01 | 0.1 | 0.76 | 0.02 | 1.13 | 1 | 1 | |
| LLaMA-3.1-8B-Instruct SFTBase Model=LLaMA-3.1-8B, Prompting Strategy=Instruct, Fine-tuning Strategy=SFT2025.04 | 0.08 | 0 | 0.05 | 0.74 | 0.01 | 1.12 | 1 | 1 |