Commonsense Reasoning on QASC (test)
90.06AccuracyRetrieval-based knowledge
Evaluation results
| Method | Links | |
|---|---|---|
| Retrieval-based knowledgeInference Model=UQA-11b-ft2021.10 | 90.06 | |
| Generated Knowledge PromptingInference Model=UQA-11b-ft2021.10 | 80.33 | |
| Vanilla baselineInference Model=UQA-11b-ft2021.10 | 76.74 | |
| UQA-11b-ft (prev. SOTA)2021.10 | 76.74 | |
| Generated Knowledge PromptingInference Model=T5-11b2021.10 | 55 | |
| Vanilla baselineInference Model=T5-11b2021.10 | 44.89 |