Natural Language Inference on NLI Suite (wanli, anli, cnli)
38.67WANLI AccuracyDOPA*
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| DOPA*LLM=GPT4o-mini2026.04 | 38.67 | 40.83 | 32.67 | 37.39 | |
| InfICLLLM=GPT4o-mini2026.04 | 38.38 | 38 | 32.83 | 36.39 | |
| DOPA*LLM=GPT3.5-turbo2026.04 | 38.17 | 39.33 | 32 | 36.5 | |
| KNNLLM=GPT3.5-turbo2026.04 | 38 | 40 | 30.83 | 36.28 | |
| KNNLLM=GPT4o-mini2026.04 | 37.67 | 39.67 | 26.83 | 34.72 | |
| InfICLLLM=GPT3.5-turbo2026.04 | 37.67 | 39.5 | 32.5 | 36.56 | |
| DICLLLM=GPT4o-mini2026.04 | 37 | 36.67 | 20.5 | 31.39 | |
| RewriteLLM=GPT3.5-turbo2026.04 | 37 | 38 | 28.67 | 34.56 | |
| RewriteLLM=GPT4o-mini2026.04 | 36.83 | 37.67 | 27.67 | 34.06 | |
| DICLLLM=GPT3.5-turbo2026.04 | 10.67 | 13.67 | 8.17 | 10.83 |