ACSci/v3-eval-judge-gpt-oss-20b

提供的README内容描述了一个包含多个配置的数据集，每个配置代表不同的模型和版本。数据集包含的特征包括paper_id、title、model、strategy、pair_id、variant_id、style、query、response、finished、response_chars、score、n_criteria、n_yes和criterion_scores。每个配置都有一个名为iclr_2026_oral的分割，指定了num_bytes和num_examples。该数据集的结构旨在评估不同条件下各种AI模型的响应。

The provided README content describes a dataset with multiple configurations, each representing different models and versions. The dataset includes features such as paper_id, title, model, strategy, pair_id, variant_id, style, query, response, finished, response_chars, score, n_criteria, n_yes, and criterion_scores. Each configuration has a split named iclr_2026_oral with specified num_bytes and num_examples. The dataset is structured for evaluating responses from various AI models under different conditions.