Legal Sentencing Prediction on CAIL 2018
0.4618AccuracyMSR2-Qwen3-8B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MSR2-Qwen3-8BMethod Category=Fine-tuning2026.02 | 0.4618 | 0.3674 | 0.317 | 0.2884 | |
| BERTMethod Category=Classification Methods2026.02 | 0.4234 | 0.2378 | 0.2351 | 0.2264 | |
| RobertaMethod Category=Classification Methods2026.02 | 0.4212 | 0.289 | 0.2414 | 0.2308 | |
| TFIDF+SVMMethod Category=Classification Methods2026.02 | 0.4205 | 0.2856 | 0.2642 | 0.2666 | |
| Qwen3-8B-FTMethod Category=Fine-tuning2026.02 | 0.4184 | 0.2957 | 0.2389 | 0.2231 | |
| CNNMethod Category=Classification Methods2026.02 | 0.3939 | 0.1938 | 0.1751 | 0.1608 | |
| Qwen3-235B-A22B-DIMethod Category=Direct Inference, Evaluation Protocol=Zero-shot2026.02 | 0.3805 | 0.2936 | 0.2855 | 0.2682 | |
| Llama3-8B-FTMethod Category=Fine-tuning2026.02 | 0.3715 | 0.2683 | 0.2137 | 0.1938 | |
| NeurJudgeMethod Category=Classification Methods2026.02 | 0.3684 | 0.2099 | 0.1681 | 0.1851 | |
| LADANMethod Category=Classification Methods2026.02 | 0.3614 | 0.2086 | 0.1803 | 0.1658 | |
| TopJudgeMethod Category=Classification Methods2026.02 | 0.357 | 0.1859 | 0.1843 | 0.1763 | |
| Qwen3-14B-RAGMethod Category=Retrieval-augmented Generation2026.02 | 0.2968 | 0.2282 | 0.21 | 0.1957 | |
| Qwen3-8B-RAGMethod Category=Retrieval-augmented Generation2026.02 | 0.2816 | 0.2054 | 0.2156 | 0.187 | |
| Qwen3-32B-DRMethod Category=Direct Reasoning2026.02 | 0.2769 | 0.2321 | 0.2657 | 0.2222 | |
| PLJP-GPT-4o-miniMethod Category=Retrieval-augmented Generation2026.02 | 0.2753 | 0.2365 | 0.1956 | 0.1803 | |
| Qwen3-32B-DIMethod Category=Direct Inference, Evaluation Protocol=Zero-shot2026.02 | 0.2649 | 0.2768 | 0.2728 | 0.2343 | |
| Qwen3-14B-DIMethod Category=Direct Inference, Evaluation Protocol=Zero-shot2026.02 | 0.2558 | 0.1935 | 0.2244 | 0.1803 | |
| Qwen3-14B-DRMethod Category=Direct Reasoning2026.02 | 0.2558 | 0.1935 | 0.2244 | 0.1803 | |
| Qwen3-8B-DRMethod Category=Direct Reasoning2026.02 | 0.2057 | 0.2043 | 0.2107 | 0.1673 | |
| Llama3-8B-DIMethod Category=Direct Inference, Evaluation Protocol=Zero-shot2026.02 | 0.1254 | 0.1491 | 0.1312 | 0.086 | |
| Qwen3-8B-DIMethod Category=Direct Inference, Evaluation Protocol=Zero-shot2026.02 | 0.1138 | 0.3192 | 0.1404 | 0.0872 |