DeepSeek-R1-Medical-COT

DeepSeek-R1-Medical-COT

DeepSeek-R1-Medical-COT is a 4-bit fine-tuned language model optimized for medical reasoning and clinical scenario interpretation.
It is based on unsloth/DeepSeek-R1-Distill-Llama-8B and fine-tuned on the FreedomIntelligence/medical-o1-reasoning-SFT dataset to provide structured, step-by-step clinical reasoning and evidence-based conclusions.

---

Model Details

• Developed by: Mohamed Adel
• Model type: Causal Language Model (LLM)
• Language: English
• License: Apache-2.0
• Base model: unsloth/DeepSeek-R1-Distill-Llama-8B
• Finetuned for: Medical instruction-following and clinical reasoning tasks

Model Sources

• Repository: Hugging Face Model Hub
• Dataset for fine-tuning: FreedomIntelligence/medical-o1-reasoning-SFT

---

Uses

Direct Use

• Answer medical questions with step-by-step reasoning
• Predict clinical outcomes from scenarios
• Assist healthcare professionals in education or training

Downstream Use

• Integrate into medical decision-support tools
• Knowledge-grounded chatbots for clinical education
• Further fine-tuning for specialized medical domains

Out-of-Scope Use

• Real-time diagnosis for patients without supervision
• Legal or financial medical advice
• Non-medical tasks

---

How to Use

from unsloth import FastLanguageModel
from transformers import AutoTokenizer

model_name = "DeepSeek-R1-Medical-COT"

# Load model
model, tokenizer = FastLanguageModel.from_pretrained(model_name, load_in_4bit=True)

# Example inference
prompt = """
### Clinical Scenario:
A 54-year-old man complains of frequent urinary urgency, nocturia, and a weak urinary stream. His prostate is moderately enlarged. Predict likely cystometric findings.
"""
inputs = tokenizer([prompt], return_tensors="pt").to("cuda")
outputs = model.generate(
    input_ids=inputs.input_ids,
    attention_mask=inputs.attention_mask,
    max_new_tokens=500
)
print(tokenizer.decode(outputs[0]))

Training Details

• Dataset: FreedomIntelligence/medical-o1-reasoning-SFT
• Preprocessing: Prompts formatted with CoT style (<think>...</think>) for step-by-step reasoning
• Fine-Tuning Method: LoRA applied to attention and feedforward modules
• Hyperparameters:
  • Batch size: 1 (gradient accumulation 8)
  • Max steps: 200
  • Learning rate: 2e-4
  • Mixed precision: FP16 / BF16 depending on GPU support
  • Optimizer: 8-bit AdamW

Evaluation

• Evaluated on a subset of medical reasoning questions
• Metrics: correctness of step-by-step reasoning, coherence, and final answer accuracy
• Results indicate improved structured reasoning over base model

Limitations and Risks

• Limited to the quality and scope of the training dataset
• May not cover rare or highly specialized medical cases
• Should not replace clinical judgment; intended for educational and reasoning support

Recommendation: Always review model outputs with a qualified healthcare professional.

Mohamed Adel (2026). DeepSeek-R1-Medical-COT. Retrieved from https://huggingface.co/DeepSeek-R1-Medical-COT