Sentencing Prediction on CAIL Law Domain
85AccuracyRe-TASK (Lite)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Re-TASK (Lite)Backbone=Yi-1.5-9B, Prompting Strategy=Re-TASK (Lite)2024.08 | 85 | 27.17 | |
| Re-TASK (Lite)Backbone=Llama3-Chinese-8B, Prompting Strategy=Re-TASK (Lite)2024.08 | 78.5 | 27.17 | |
| Step-BackBackbone=Llama3-Chinese-8B, Prompting Strategy=Step-Back2024.08 | 72.5 | 18 | |
| Step-BackBackbone=Yi-1.5-9B, Prompting Strategy=Step-Back2024.08 | 72.5 | 18 | |
| Three-shot CoTBackbone=Yi-1.5-9B, Prompting Strategy=Three-shot CoT2024.08 | 70.17 | 12.5 | |
| One-shot CoTBackbone=Yi-1.5-9B, Prompting Strategy=One-shot CoT2024.08 | 66.5 | 9.61 | |
| Re-TASK (+K0)Backbone=Llama3-Chinese-8B, Prompting Strategy=Re-TASK (+K0)2024.08 | 60.5 | 11.5 | |
| Re-TASK (+K0)Backbone=Yi-1.5-9B, Prompting Strategy=Re-TASK (+K0)2024.08 | 57.5 | 11.5 | |
| Three-shot CoTBackbone=Llama3-Chinese-8B, Prompting Strategy=Three-shot CoT2024.08 | 56.33 | 12.5 | |
| Zero-shot CoT + SCBackbone=Llama3-Chinese-8B, Prompting Strategy=Zero-shot CoT + SC2024.08 | 54.5 | 0.33 | |
| Plan-and-SolveBackbone=Llama3-Chinese-8B, Prompting Strategy=Plan-and-Solve2024.08 | 54.5 | 1.83 | |
| Zero-shot CoTBackbone=Llama3-Chinese-8B, Prompting Strategy=Zero-shot CoT2024.08 | 54 | — | |
| One-shot CoTBackbone=Llama3-Chinese-8B, Prompting Strategy=One-shot CoT2024.08 | 53.67 | 9.61 | |
| Re-TASK (Lite)Backbone=Qwen1.5-7B, Prompting Strategy=Re-TASK (Lite)2024.08 | 45.5 | 27.17 | |
| Plan-and-SolveBackbone=Qwen1.5-7B, Prompting Strategy=Plan-and-Solve2024.08 | 45 | 1.83 | |
| Re-TASK (+K0)Backbone=Qwen1.5-7B, Prompting Strategy=Re-TASK (+K0)2024.08 | 44 | 11.5 | |
| Zero-shot CoT + SCBackbone=Yi-1.5-9B, Prompting Strategy=Zero-shot CoT + SC2024.08 | 40.5 | 0.33 | |
| Zero-shot CoTBackbone=Yi-1.5-9B, Prompting Strategy=Zero-shot CoT2024.08 | 40 | — | |
| Three-shot CoTBackbone=Qwen1.5-7B, Prompting Strategy=Three-shot CoT2024.08 | 38.5 | 12.5 | |
| Step-BackBackbone=Qwen1.5-7B, Prompting Strategy=Step-Back2024.08 | 36.5 | 18 | |
| One-shot CoTBackbone=Qwen1.5-7B, Prompting Strategy=One-shot CoT2024.08 | 36.17 | 9.61 | |
| Zero-shot CoTBackbone=Qwen1.5-7B, Prompting Strategy=Zero-shot CoT2024.08 | 33.5 | — | |
| Zero-shot CoT + SCBackbone=Qwen1.5-7B, Prompting Strategy=Zero-shot CoT + SC2024.08 | 33.5 | 0.33 | |
| Plan-and-SolveBackbone=Yi-1.5-9B, Prompting Strategy=Plan-and-Solve2024.08 | 33.5 | 1.83 |