meno_llm_3b_v2

meno_llm_3b_v2

This repository contains a LoRA adapter fine-tuned on top of:

Base model: hishab/titulm-llama-3.2-3b-v2.0

Description

This model is being developed as a Bangla empathetic conversational assistant for health-related question answering.

It is intended to be used in a retrieval-augmented generation (RAG) pipeline, where relevant context is retrieved first and then passed to the model for response generation.

Important Note

This repository contains only the LoRA adapter, not the full standalone base model.

To use it, load it on top of:

hishab/titulm-llama-3.2-3b-v2.0

Current Status

• Fine-tuned adapter uploaded successfully
• Produces empathetic Bangla answers
• Works better when used with retrieved supporting context
• Still under improvement for grounding, medical faithfulness, and weak-context handling

Intended Use

• Bangla conversational response generation
• Empathetic health-related QA
• RAG-based assistant workflows
• Research and development use

Limitations

• This model is still under active development
• It may produce incorrect or weakly grounded answers if retrieval quality is poor
• It should not be treated as a substitute for professional medical advice

Loading

Use this adapter together with the base TiTuLM model.

Base model: hishab/titulm-llama-3.2-3b-v2.0

Adapter repo: afifaimran/meno_llm_3b_v2

Author

Afifa Imran