VILA-Lab/DRAG

(ACL 2025 Main) Distilling RAG for SLMs from LLMs to Transfer Knowledge and Mitigate Hallucination via Evidence and Graph-based Distillation

22
/ 100
Experimental

This project helps AI developers create smaller, more efficient Retrieval-Augmented Generation (RAG) models. It takes knowledge and reasoning abilities from large, powerful language models (LLMs) and transfers them into smaller language models (SLMs). The output is an SLM that can answer factual questions more accurately and with less "hallucinated" content, while significantly reducing computational costs and model size.

No commits in the last 6 months.

Use this if you need to deploy a RAG system that is factually consistent and computationally lightweight, without sacrificing accuracy.

Not ideal if you are a business user looking for a ready-to-use RAG application, as this is a framework for AI model development.

AI-model-distillation factual-AI LLM-optimization hallucination-mitigation knowledge-transfer
No License Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 7 / 25
Maturity 7 / 25
Community 6 / 25

How are scores calculated?

Stars

34

Forks

2

Language

Python

License

Last pushed

Aug 23, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/rag/VILA-Lab/DRAG"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.