Natural Language Inference on EntailmentBank (test)
54BLEUllama3-70b-8192
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| llama3-70b-8192Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=102026.01 | 54 | 0.97 | 54 | |
| Flan-T5 (xxl)Architecture=encoder-decoder, Inference Type Injection (INJ)=Yes, Num examples=52026.01 | 53 | 0.97 | 43 | |
| GPT-4-0613Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=102026.01 | 53 | 0.97 | 50 | |
| GPT-3.5-turbo-0125Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=102026.01 | 52 | 0.96 | 40 | |
| GPT-4-0613Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=52026.01 | 52 | 0.97 | 47 | |
| GPT-4-0613Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=02026.01 | 52 | 0.97 | 50 | |
| llama3-70b-8192Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=52026.01 | 52 | 0.97 | 52 | |
| CoT-T5 (11b)Architecture=encoder-decoder, Inference Type Injection (INJ)=Yes, Num examples=102026.01 | 51 | 0.97 | 39 | |
| CoT-T5 (11b)Architecture=encoder-decoder, Inference Type Injection (INJ)=Yes, Num examples=52026.01 | 51 | 0.97 | 39 | |
| Flan-T5 (xxl)Architecture=encoder-decoder, Inference Type Injection (INJ)=Yes, Num examples=102026.01 | 51 | 0.97 | 41 | |
| llama3-70b-8192Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=02026.01 | 51 | 0.97 | 47 | |
| CoT-T5 (11b)Architecture=encoder-decoder, Inference Type Injection (INJ)=Yes, Num examples=02026.01 | 50 | 0.97 | 36 | |
| Flan-T5 (xxl)Architecture=encoder-decoder, Inference Type Injection (INJ)=Yes, Num examples=02026.01 | 50 | 0.96 | 37 | |
| Flan-T5 (xxl)Architecture=encoder-decoder, Inference Type Injection (INJ)=No, Num examples=02026.01 | 48 | 0.96 | 36 | |
| GPT-3.5-turbo-0125Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=52026.01 | 48 | 0.96 | 35 | |
| GPT-4-0613Architecture=decoder only, Inference Type Injection (INJ)=No, Num examples=02026.01 | 47 | 0.96 | 40 | |
| CoT-T5 (11b)Architecture=encoder-decoder, Inference Type Injection (INJ)=No, Num examples=02026.01 | 46 | 0.96 | 31 | |
| GPT-3.5-turbo-0125Architecture=decoder only, Inference Type Injection (INJ)=Yes, Num examples=02026.01 | 46 | 0.96 | 31 | |
| llama3-70b-8192Architecture=decoder only, Inference Type Injection (INJ)=No, Num examples=02026.01 | 44 | 0.96 | 40 | |
| GPT-3.5-turbo-0125Architecture=decoder only, Inference Type Injection (INJ)=No, Num examples=02026.01 | 42 | 0.96 | 33 |