On Verbalized Confidence Scores for LLMs
收藏官方服务:
资源简介:
We provide all raw responses and predictions from our evaluated LLMs. The results.zip consists of the following directories: responses: contains all raw responses in JSON format predictions: contains all predicted confidence scores of valid responses and the corresponding ground-truth correctness labels in numpy format predictions_sampled: contains 1000 random samples (with replacement) from predictions used for evaluation in our paper
提供机构:
Zenodo创建时间:
2024-12-19



