Medical-Reasoning-SFT-GPT-OSS-120B-Small

# Medical-Reasoning-SFT-GPT-OSS-120B-Small A filtered and processed version of [OpenMed/Medical-Reasoning-SFT-GPT-OSS-120B](https://huggingface.co/datasets/OpenMed/Medical-Reasoning-SFT-GPT-OSS-120B) optimized for training efficiency. ## Dataset Description This dataset contains high-quality medical reasoning conversations with the following modifications: 1. **Length Filtering**: Only includes samples where assistant responses are between 1000 and 10000 characters 2. **Reasoning Extraction**: Reasoning content from `<think>` tags has been extracted into a separate `reasoning_content` field ## Statistics - **Original dataset size**: 200,193 samples - **Filtered dataset size**: 131,728 samples (65.8%) - **Samples with reasoning**: 131,728 - **Samples without reasoning**: 0 ## Filtering Criteria - **Minimum assistant response length**: 1000 characters - **Maximum assistant response length**: 10000 characters - **Removed**: 13,664 samples (too short) + 54,801 samples (too long) ## Dataset Structure Each sample contains: - `messages`: List of conversation messages with roles (user/assistant) - Assistant messages now have two fields: - `content`: The actual response (without `<think>` tags) - `reasoning_content`: The extracted reasoning from `<think>` tags ## Usage ```python from datasets import load_dataset dataset = load_dataset("OpenMed/Medical-Reasoning-SFT-GPT-OSS-120B-Small") # Access a sample sample = dataset['train'][0] messages = sample['messages'] for msg in messages: if msg['role'] == 'assistant': reasoning = msg.get('reasoning_content', '') response = msg.get('content', '') print(f"Reasoning: {reasoning}") print(f"Response: {response}") ``` ## Source Dataset This dataset is derived from [OpenMed/Medical-Reasoning-SFT-GPT-OSS-120B](https://huggingface.co/datasets/OpenMed/Medical-Reasoning-SFT-GPT-OSS-120B), which was generated using OpenAI's gpt-oss-120B model. ## License Apache 2.0 (inherited from source dataset) ## Citation If you use this dataset, please cite the original dataset: ``` @dataset{medical_reasoning_sft_gpt_oss_120b, title={Medical-Reasoning-SFT-GPT-OSS-120B}, author={OpenMed}, year={2024}, publisher={Hugging Face}, url={https://huggingface.co/datasets/OpenMed/Medical-Reasoning-SFT-GPT-OSS-120B} } ```