Kiria-Nozan/TRIM-gpt-oss-120b-separate-neighbors-only-para-random-feature-num

该目录是TRIM代理推理SFT数据的Hugging Face友好型公共导出。数据集包括训练分割，包含10056条记录，涉及多个任务，如AMES、BBB_Martins、Bioavailability_Ma等。每条记录是一个JSONL格式的训练示例，包含多个字段，如schema_version、sft_mode、task、split、sample_index、sample_id、smiles、gt_label、final_answer_option和messages。messages字段存储了工具增强的聊天记录，包括嵌套的tool_calls和原始SFT导出中使用的助手thinking文本。数据集已进行公共清理，移除了本地绝对source_paths。

This directory is a Hugging Face-friendly public export of the TRIM agent reasoning SFT data. The dataset includes a train split with 10056 records, covering multiple tasks such as AMES, BBB_Martins, Bioavailability_Ma, etc. Each record is a JSONL-formatted training example with fields like schema_version, sft_mode, task, split, sample_index, sample_id, smiles, gt_label, final_answer_option, and messages. The messages field stores a tool-augmented chat transcript, including nested tool_calls and the assistant thinking text used in the original SFT export. The dataset has been sanitized for public use, with local absolute source_paths removed.