Rumiii/LiquiMedThink1.2B

LiquiMedThink 1.2B: Medical Reasoning with Explicit Thinking

LiquiMedThink 1.2B is a specialized 1.17 billion parameter model, fine-tuned from LiquidAI/LFM2.5-1.2B-Thinking, designed for medical chain-of-thought reasoning. Its core innovation lies in its ability to produce explicit, step-by-step thinking traces before delivering a concise final answer, a behavior successfully preserved through a carefully structured fine-tuning process.

Key Capabilities & Features

• Retained Reasoning: Unlike prior attempts, this model successfully maintains the base model's 'thinking out loud' capability, producing distinct reasoning steps and final answers.
• Medical Domain Knowledge: Fine-tuned on the FreedomIntelligence/medical-o1-reasoning-SFT dataset (19,704 samples), it integrates clinical knowledge for medical question answering.
• Structured Output: Generates output with clear <think> tags for the reasoning process and a separate final response, mirroring the training data format.
• Small Model Efficiency: At 1.2B parameters, it offers a compact solution for exploring complex reasoning in a specialized domain.
• 4096 Token Context: Supports longer medical case descriptions and detailed reasoning traces.

What Makes This Model Different?

This model is the result of a research project demonstrating that dataset format is critical for retaining reasoning abilities during domain-specific fine-tuning. It uses a dataset where reasoning traces and final answers are genuinely distinct content, preventing the loss or corruption of the model's thinking behavior.

Intended Use Cases

• Research: Investigating small LLM capabilities in healthcare and clinical reasoning.
• Medical Education: Demonstrating clinical reasoning processes.
• Prototyping: Evaluating medical AI pipelines.
• Study: Analyzing the impact of fine-tuning data format on reasoning.

Note: This model is for research and educational purposes only and is not intended for clinical decision support or real patient care.