This model is a 7-billion parameter causal transformer fine-tuned from Qwen/Qwen2.5-7B-Instruct using Direct Preference Optimization (DPO) and Low-Rank Adaptation (LoRA). It has been aligned to generate multi-turn Socratic dialogues rather than direct factual answers, explicitly designed to foster higher-order thinking in K-12 learners.

Developed by: Mudit Jain
Model type: Causal transformer (instruction-tuned)
Language(s): English
License: MIT
Finetuned from: Qwen/Qwen2.5-7B-Instruct
Hosted on: Hugging Face (https://huggingface.com/mudit23/Socratic-Qwen2.5-7B-v2)
Dataset used: mudit23/class7-socratic-dpo

Uses

Direct Use

Generate Socratic prompts to guide students through inquiry-based learning.
Integrate into chatbots or tutoring platforms to encourage critical thinking.

Out-of-Scope Use

Do not use for tasks requiring direct factual answers or high-stakes decision making.
Not suitable for medical, legal, or safety-critical applications.

Bias, Risks, and Limitations

Hallucinations: May produce plausible-sounding but incorrect guidance.
Bias: Trained on NCERT Science material and general web data; may reflect cultural or curricular biases.
Engagement: Designed for educational contexts; performance may degrade outside Class 7 Science or similar domains.

Recommendations

Always supervise student interactions; verify any factual claims.
Restrict use to guided learning environments with educator oversight.

How to Get Started with the Model

python

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline

tokenizer = AutoTokenizer.from_pretrained("mudit23/Socratic-Qwen2.5-7B-v2")
model = AutoModelForCausalLM.from_pretrained("mudit23/Socratic-Qwen2.5-7B-v2")
generator = pipeline("text-generation", model=model, tokenizer=tokenizer)

prompt = "Teacher: What do you think causes wind to blow?"
print(generator(prompt, max_new_tokens=100)[0]["generated_text"])

Citation

BibTeX: @misc{jain2025socraticdpo, title={Enhancing K-12 Critical Thinking through a DPO-Fine-Tuned Socratic Multi-Agent AI Tutor}, author={Jain, Mudit}, year={2025}, howpublished={\url{https://huggingface.com/mudit23/Socratic-Qwen2.5-7B-v2}} }