List-based question answering on Wiki-Category list Harder (test)
22PrecisionCoVe (factored)
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CoVe (factored)LLM=Llama 65B2023.09 | 22 | 52 | 1.52 | |
| CoVe (two-step)LLM=Llama 65B2023.09 | 21 | 50 | 0.52 | |
| CoVe (joint)LLM=Llama 65B2023.09 | 15 | 30 | 1.69 | |
| Few-shotLLM=Llama 65B2023.09 | 12 | 55 | 4.05 | |
| Zero-shotLLM=Llama 2 70B Chat2023.09 | 5 | 35 | 6.85 | |
| CoTLLM=Llama 2 70B Chat2023.09 | 3 | 30 | 11.1 |