DoRAG: The Retrieval Augmented Generation Challenge

Calling all AI enthusiasts!

Are you ready to push the boundaries of Retrieval Augmented Generation (RAG) models? DoRAG is your chance to showcase your skills and build a state-of-the-art RAG model within a thrilling 6-hour time limit (5 PM to 11 PM).

What is DoRAG?

DoRAG is a fast-paced competition where participants will race against the clock to develop and train a RAG model. You'll be provided with a rich dataset and access to powerful computing resources. Your task is to build the most effective RAG model within the timeframe, focusing on:

• Accuracy: How well does your model answer questions or complete tasks using retrieved information?
• Retrieval Relevance: Do the retrieved documents accurately support the generated response?
• Efficiency: How efficiently can your model retrieve relevant information?

What do you need?

• A team of passionate AI enthusiasts (2 - 3/team)
• A thirst for competition and innovation

What will you win?

• Bragging rights as the DoRAG champion!
• Exciting prizes (to be announced!)

Ready to DoRAG?

Registration for DoRAG opens on 2024-05-20. Don't miss this opportunity to test your skills and contribute to the advancement of RAG technology.

Rules:

• Team Size (2 - 3)
• All participants in a team should be from the same department, but can be from different year of study
• Submission should be made as a link to GitHub Repo
• Should only use Open-Source models to build RAG
• Allowed to use ChatGPT / GEMINI

Evaluation Criteria:

Basic:

• File and Code structure
• Open-sourcability of project
• UI Design scale and frontend features

Advanced:

• Retrieval Relevance
• RAG pipeline Novelty
• Document types supported

There will also be hidden criteria which will be shared only post submission.