DCAgent3/aider_polyglot_a3_rl_DCAgent_inferredbugs_sandboxes_verifier_55_8B_20260526_230125
收藏资源简介:
--- dataset_info: features: - name: conversations list: - name: role dtype: string - name: content dtype: string - name: agent dtype: string - name: model dtype: string - name: model_provider dtype: string - name: date dtype: string - name: task dtype: string - name: episode dtype: string - name: run_id dtype: string - name: trial_name dtype: string - name: result dtype: string - name: verifier_output dtype: string - name: trace_source dtype: string splits: - name: train num_bytes: 57689010 num_examples: 869 download_size: 51122130 dataset_size: 57689010 configs: - config_name: default data_files: - split: train path: data/train-* ---
This dataset consists of multi-turn conversation records and related metadata, designed for analyzing and evaluating the interactive behaviors of AI agents or models. Features include conversations (with role and content fields), agent identifier, model information (model and model_provider), date, task type, episode number, run ID, trial name, result, verifier output, and trace source. The dataset contains 869 training examples with a total size of approximately 57.7 MB, suitable for natural language processing tasks such as dialogue system evaluation, agent behavior analysis, or model performance research.




