You can deploy this model and start earning money today!

RohitFrancis/
DeepSeek-R1-Interactive-Reasoning-finetune

Name: DeepSeek-R1-Interactive-Reasoning-finetune
Availability: InStock

gguf

endpoints_compatible

conversational

Views

No views yet

DeepSeek-R1-Interactive-Reasoning (Reasoning + Clarification)

Model Summary

DeepSeek-R1-Interactive-Reasoning is a fine-tuned reasoning model based on DeepSeek-R1. Unlike standard reasoning models that only output final answers, this model explicitly integrates clarification tokens into its reasoning process:

<ask> → requests missing context from the user.
<question> → asks for specific factual input needed for reasoning.
<doubt> → expresses uncertainty, highlighting assumptions or limitations.

This makes the model more interactive, transparent, and safe for ambiguous queries.

Intended Uses

Interactive reasoning in education, research, planning, and assistance systems.
Applications where ambiguity must be clarified before giving a final answer.
Safer deployment in healthcare, legal, or customer support where incorrect assumptions can be harmful.

Not intended for:

Replacing domain experts (medical, legal, financial advice).
Unmoderated use where clarifications may be ignored.

Training Details

Base Model: DeepSeek-R1
Method: Supervised fine-tuning (LoRA/Full SFT)
Dataset Size: ~5,000 curated examples
Dataset Type: Reasoning traces with <ask>, <doubt>, <question> tags

Example Usage

first run ollama create modelname_of_your_choice -f ./Modelfile

python

import ollama

response = ollama.chat(
    model="<The Model Name you gave>",
    messages=[
        {"role": "user", "content": "I have fever and vomiting, what is happening to me?"}
    ]
)

print(response.message.content)

Output:

You are experiencing fever and vomiting, which could indicate many different conditions. 
<question>What is your temperature?</question>
(User responds: 102°F)
Okay, you have a high fever, and you are vomiting. These are classic symptoms of flu or other viral infections. 
<doubt>I am not a doctor, but the most common thing with fever and vomiting at 102°F is likely a viral infection.</doubt>
Final Answer: Most likely you have a viral infection such as the flu. However, consult a medical professional for an accurate diagnosis.

Another Example

User: "I want to plan an industrial visit with my engineering college."

Model Output:

Okay, to plan a successful industrial visit, I need some details. 
<question>What industry are you interested in? For example, IT, Automobile, Robotics, etc.</question>
(User responds: Robotics)
<ask>What is the budget for this visit? Including travel expenses, entrance fees and other costs?</ask>
(User responds: 20000 INR)
<ask>Which companies do you want to visit that are in the robotics space? Are they nearby or do you need to travel far?</ask>
(User responds: Boston Robotics Inc. in Bangalore)
Final Answer: Plan an industrial visit to Boston Robotics Inc, Bangalore with a budget of 20,000 INR, focusing on robotics.

Limitations

May overuse clarification tokens even when unnecessary.
Only trained on ~5k samples, so robustness on out-of-domain queries may be limited.
Not a medical or legal advisor. Outputs should be used for informational purposes only.

Ethical Considerations

Designed to reduce hallucination risk by prompting for clarification.
Still requires human judgment before acting on outputs.
Explicit <doubt> tokens highlight uncertainty — but users must not ignore them.

Future Work

Expand dataset beyond 5k examples for stronger generalization.
Collect preference data to fine-tune when clarifications are appropriate.
Add multilingual clarification reasoning.

Citation

If you use this model, please cite:

@misc{deepseek-rc,
  title        = {DeepSeek-RC: Reasoning + Clarification Model},
  author       = {Your Name},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/RohitFrancis/DeepSeek-R1-Interactive-Reasoning-finetune/}}
}