mugezhang/medical-temporal-reasoning-sft

--- configs: - config_name: default data_files: - split: train_answer_only path: sft_train_answer_only.json - split: train_reasoning path: sft_train_reasoning.json - split: val_answer_only path: sft_val_answer_only.json - split: val_reasoning path: sft_val_reasoning.json features: - name: item_id dtype: string - name: image_path sequence: string - name: question dtype: string - name: answer dtype: string - name: process dtype: string - name: dataset_name dtype: string --- # Medical Temporal Reasoning — SFT Dataset SFT training data for a medical temporal reasoning model that, given a current and prior chest X-ray, produces step-by-step reasoning about disease progression. Built from MIMIC-CXR image pairs. Images are not included; `image_path` contains relative paths into the MIMIC-CXR-JPG dataset. ## Splits | Split | Records | Description | |---|---|---| | `train_answer_only` | 158,439 | Training records, answer only (`process=null`) | | `train_reasoning` | 10,559 | Training records with GPT-mined `<think>` traces | | `val_answer_only` | 7,933 | Validation records, answer only (`process=null`) | | `val_reasoning` | 516 | Validation records with `<think>` traces | The four splits are fully disjoint (no shared `item_id`). Train/val split is by patient (`subject_id`) — no patient appears in both. ## Sources | dataset_name | Type | Records | |---|---|---| | `mmxu` | MCQ, region-level progression | 90,000 | | `meddiffvqa` | Open-ended, entity-level change | 85,339 | | `custom_mcq_ms_cxr_t` | MCQ, disease-level (GPT-generated from MS-CXR-T) | 2,108 | ## Format Each record: ```json { "item_id": "mmxu_12345", "image_path": ["p17/.../current.jpg", "p17/.../prior.jpg"], "question": "How has the left lung changed? A) Worsened B) Improved ...", "answer": "<answer>A</answer>", "process": "<think>Comparing the two images...</think>", "dataset_name": "mmxu_answer_only" } ``` `process` is `null` in answer_only splits and a `<think>...</think>` string in reasoning splits.配置项： - 配置名称：default 数据文件： - 划分：train_answer_only，路径：sft_train_answer_only.json - 划分：train_reasoning，路径：sft_train_reasoning.json - 划分：val_answer_only，路径：sft_val_answer_only.json - 划分：val_reasoning，路径：sft_val_reasoning.json 特征字段： - 名称：item_id，数据类型：字符串 - 名称：image_path，数据类型：字符串序列 - 名称：question，数据类型：字符串 - 名称：answer，数据类型：字符串 - 名称：process，数据类型：字符串 - 名称：dataset_name，数据类型：字符串 # 医疗时序推理——监督微调（Supervised Fine-Tuning, SFT）数据集 本数据集为医疗时序推理模型的监督微调训练数据，该模型可基于当前及既往胸部X线影像，生成关于疾病进展的分步推理过程。数据集基于MIMIC-CXR影像对构建，未包含原始影像；`image_path`字段包含指向MIMIC-CXR-JPG数据集的相对路径。 ## 数据集划分 | 划分 | 样本量 | 说明 | |---|---|---| | `train_answer_only` | 158,439 | 仅包含答案的训练样本，`process`字段为`null` | | `train_reasoning` | 10,559 | 包含GPT挖掘的`<think>`推理轨迹的训练样本 | | `val_answer_only` | 7,933 | 仅包含答案的验证样本，`process`字段为`null` | | `val_reasoning` | 516 | 包含`<think>`推理轨迹的验证样本 | 四个数据集划分完全互斥（无重复的`item_id`），训练/验证划分基于患者（`subject_id`）进行划分——无患者同时出现在训练集与验证集中。 ## 数据源 | 数据集名称 | 任务类型 | 样本量 | |---|---|---| | `mmxu` | 多项选择题（Multiple Choice Question, MCQ），区域级进展分析 | 90,000 | | `meddiffvqa` | 开放式问答，实体级变化分析 | 85,339 | | `custom_mcq_ms_cxr_t` | 多项选择题（MCQ），疾病级进展分析（由GPT基于MS-CXR-T生成） | 2,108 | ## 数据格式 每条样本的JSON结构如下： json { "item_id": "mmxu_12345", "image_path": ["p17/.../current.jpg", "p17/.../prior.jpg"], "question": "左侧肺部有何变化？A) 恶化 B) 改善 ...", "answer": "<answer>A</answer>", "process": "<think>对比两张影像……</think>", "dataset_name": "mmxu_answer_only" } 其中，仅答案划分的样本中`process`字段为`null`，带推理划分的样本中`process`为`<think>...</think>`格式的推理字符串。