Anonymiserad/SweMedMCQ-LFM2.5-1.2B-Trace-EN-SV
Anonymiserad/SweMedMCQ-LFM2.5-1.2B-Trace-EN-SV is a 1.2 billion parameter language model, fine-tuned from LiquidAI/LFM2.5-1.2B-Instruct, specifically for medical multiple-choice question answering in both English and Swedish. It utilizes a reasoning-trace supervision method, where it learns from detailed explanations provided by larger teacher models. This model excels at generating structured JSON responses with explanations for medical questions, making it suitable for research in medical AI and benchmarking LLMs for specialized domains.
Loading preview...
SweMedMCQ-LFM2.5-1.2B-Trace-EN-SV Overview
This model is a 1.2 billion parameter language model, fine-tuned by Anonymiserad from the LiquidAI/LFM2.5-1.2B-Instruct base model. Its primary focus is on medical multiple-choice question answering in both English and Swedish. A key differentiator is its training methodology, which incorporates reasoning-trace supervision. This means the model learns not just the correct answer, but also the detailed thought process and explanation leading to that answer, derived from larger teacher models like GPT-OSS-120B and Qwen3.6-35B-A3B.
Key Capabilities
- Bilingual Medical QA: Specialized in answering medical multiple-choice questions in both English and Swedish.
- Reasoning-Trace Generation: Capable of producing a
<think>...</think>block detailing its reasoning process before outputting a structured JSON answer with an explanation. - Structured Output: Designed to return answers in a specific JSON format, including an
explanationandanswer_index. - Improved Medical Accuracy: Achieves significantly higher accuracy on the SweMedMCQ benchmark (50.5% English, 52.0% Swedish for reasoning decoding) compared to its base model (37.2% English, 23.5% Swedish).
Good For
- Medical AI Research: Ideal for researchers exploring specialized medical language models and reasoning capabilities.
- Benchmarking: Suitable for evaluating LLMs on medical question-answering tasks, particularly with the SweMedMCQ benchmark.
- Developing Explainable AI: Useful for applications requiring not just an answer, but also a transparent explanation of the reasoning behind it in a medical context.
- Bilingual Medical Information Retrieval: Can be used in scenarios requiring medical knowledge extraction and explanation in both English and Swedish.