Compositional Reasoning on Compositional Reasoning Correction Input Ic
95.3Event Probability RatioCREME
Evaluation results
| Method | Links | |
|---|---|---|
| CREMEBackbone=OpenAlpaca-3B2024.02 | 95.3 | |
| CREMEBackbone=LLAMA-2-7B2024.02 | 87.5 | |
| Memory InjectionBackbone=LLAMA-2-7B2024.02 | 63 | |
| OriginalBackbone=LLAMA-2-7B2024.02 | 59.5 | |
| Memory InjectionBackbone=OpenAlpaca-3B2024.02 | 58.7 | |
| OriginalBackbone=OpenAlpaca-3B2024.02 | 58 | |
| CoT-PatchScopesBackbone=OpenAlpaca-3B2024.02 | 57.3 | |
| CoT-PatchScopesBackbone=LLAMA-2-7B2024.02 | 53 |