Chain-of-Thought Generation on ANLI (test)
4.01GPT-4 ScoreGold
Evaluation results
| Method | Links | |
|---|---|---|
| GoldSelf-consistency=true2024.03 | 4.01 | |
| GoldSelf-consistency=false2024.03 | 3.82 | |
| DSSSelf-consistency=false2024.03 | 3.48 | |
| DSSSelf-consistency=true2024.03 | 3.44 | |
| MI-based distillationSelf-consistency=false2024.03 | 3.42 | |
| MI-based distillationSelf-consistency=true2024.03 | 3.34 |