Bio-Medical-3B-CoT-012025

This model is a fine-tuned version of Qwen2.5-3b-Instruct on our custom "BioMedData" dataset, enhanced with chain-of-thought prompting instructions to introduce advanced reasoning capabilities. It has been specifically optimized for applications in the Healthcare & Lifesciences (HLS) domain.

Model details

Model Name: Bio-Medical-3B-CoT-012025

Base Model: Qwen2.5-3b-Instruct

Parameter Count: 3 billion

Training Data: Custom high-quality biomedical dataset with chain-of-thought examples.

Number of Entries in Dataset: 600,000+

Dataset Composition: The dataset comprises both synthetic and manually curated samples, ensuring diverse and comprehensive coverage of biomedical knowledge.

Model description

The Bio-Medical-3B-CoT-012025 model is designed to provide accurate, context-aware, and reasoning-driven text generation in the biomedical domain. It has been fine-tuned on a dataset that includes chain-of-thought prompting to enable logical reasoning and better interpretability of its outputs.

This model is tailored for:

• Understanding and generating domain-specific content in the healthcare and biomedical fields.
• Answering complex questions that require step-by-step reasoning.
• Supporting professionals, researchers, and students in clinical and scientific tasks.

Evaluation Metrics

Bio-Medical-3B-CoT-012025 has been evaluated using the Eleuther AI Language Model Evaluation Harness framework on the following tasks:

• medmcqa
• medqa_4options
• mmlu_anatomy
• mmlu_clinical_knowledge
• mmlu_college_biology
• mmlu_college_medicine
• mmlu_medical_genetics
• mmlu_professional_medicine
• pubmedqa

Results show consistent performance improvements over general-purpose models of similiar size, particularly in tasks requiring reasoning.

Intended uses & limitations

Intended Uses:

1. Research Support: Assisting researchers in extracting and generating insights from biomedical texts.
2. Clinical Decision Support: Aiding in the interpretation of clinical data and evidence-based recommendations.
3. Educational Tool: Enabling students and professionals to understand complex biomedical concepts.

Limitations and Ethical Considerations:

• Biases: The model may reflect biases present in its training data. While efforts were made to mitigate biases, some may persist.
• Accuracy: The model's responses should be validated against reliable sources, especially in critical or clinical contexts.
• Ethical Use: The model is intended to complement, not replace, expert judgment. It should be deployed responsibly in high-stakes environments.

How to use

import transformers
import torch

model_id = "ContactDoctor/Bio-Medical-3B-CoT-012025"

pipeline = transformers.pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device_map="auto",
)

messages = [
    {"role": "system", "content": "You are an expert trained on healthcare and biomedical domain!"},
    {"role": "user", "content": "What are the potential causes of chronic fatigue in a 40-year-old male?"},
]

prompt = pipeline.tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

terminators = [
    pipeline.tokenizer.eos_token_id,
    pipeline.tokenizer.convert_tokens_to_ids("<|eot_id|>")
]

outputs = pipeline(
    prompt,
    max_new_tokens=256,
    eos_token_id=terminators,
    do_sample=True,
    temperature=0.6,
    top_p=0.9,
)
print(outputs[0]["generated_text"][len(prompt):])

License

This model is licensed under the Bio-Medical-3B-CoT-012025 (Non-Commercial Use Only). Please review the terms and conditions before using the model.

Contact Information

For further information, inquiries, or issues related to Bio-Medical-3B-CoT-012025, please contact:

Email: [email protected]

Website: https://www.contactdoctor.in

Training hyperparameters

The following hyperparameters were used during training:

• Learning Rate: 0.0002
• Train Batch Size: 12
• Eval Batch Size: 8
• Seed: 42
• Gradient Accumulation Steps: 4
• Total Train Batch Size: 32
• Optimizer: Adam with betas=(0.9, 0.999) and epsilon=1e-08
• LR Scheduler Type: Cosine
• LR Scheduler Warmup Ratio: 0.03
• Training Steps: 2000
• Mixed Precision Training: Native AMP

Framework versions

• PEFT: 0.11.0
• Transformers: 4.40.2
• Pytorch: 2.1.2
• Datasets: 2.19.1
• Tokenizers: 0.19.1

Citation

If you use Bio-Medical-3B-CoT-012025 in your research or applications, please cite it as follows:

@misc{ContactDoctor_Bio-Medical-3B-CoT-012025,
  author = {ContactDoctor},
  title = {Bio-Medical-3B-CoT-012025: A High-Performance Biomedical Language Model with Reasoning Capabilities},
  year = {2025},
  howpublished = {https://huggingface.co/ContactDoctor/Bio-Medical-3B-CoT-012025},
}