Ayush-Singh/reward-bench-chatgpt-4o-latest-yes-no

该数据集包含了多个子集，每个子集都针对不同的任务或场景。具体包括：易于处理的mt_bench_easy，难度较高的mt_bench_hard，涉及编程语言如hep_rust、hep_python等，以及涉及不同类型拒绝响应的refusals_offensive和refusals_dangerous等。每个子集都包含提示(prompt)、选择(chosen)、拒绝(rejected)等信息，以及相应的概率和模型信息。

The dataset consists of multiple subsets, each tailored for different tasks or scenarios. This includes mt_bench_easy for easy processing, mt_bench_hard for more challenging tasks, programming language related subsets like hep_rust, hep_python, and subsets for different types of refusal responses such as refusals_offensive and refusals_dangerous. Each subset contains information such as prompts, choices, rejections, and corresponding probabilities and model information.