EssentialAI/cruxeval_i_adv
收藏官方服务:
资源简介:
CruxEval-Input Adversarial是一个评估模型在链式推理中检测和纠正错误能力的对抗性数据集,包含605个实例,每个实例由代码、输入、输出、正确推理链和对抗性推理链组成。
CruxEval-Input Adversarial is an adversarial dataset designed to evaluate a models ability to detect and correct reasoning errors in chain-of-thought (CoT) prompts, containing 605 instances, each comprising code, input, output, correct CoT, and adversarial CoT.
提供机构:
EssentialAI


