Retrieval-Augmented Generation Improves LLM Performance on Competitive Programming
Original title: Assessing the Impact of Retrieval-Augmented Generation on Algorithmic Problem-Solving
Embedding-based retrieval with similarity search enhances LLMs in solving algorithmic problems from CodeContests dataset.
Researchers develop RAG approach for LLMs in competitive programming. Integrates embedding retrieval and dynamic threshold to mitigate noise. Evaluates on CodeContests test split with similar training instances. Shows improved pass rates and Elo ratings. Min-K analysis distinguishes generalization from memorization.
- RAG method uses embedding-based retrieval for similar problems
- Dynamic threshold mitigates retrieval noise
- Evaluated on CodeContests dataset test split
- Measures performance with pass rates and CodeForces Elo
From the original article
1 min read at arxiv.orgCompetitive programming problems pose distinctive challenges for Large Language Models, as they demand not only proficient code generation but also advanced algorithmic reasoning. This study presents a Retrieval-Augmented Generation approach specifically designed to enhance the capacity of Large Language Models to address such problems.
By Gabriel Souza Baggio · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.