LLMs can learn self-restraint through iterative self-reflection
Welcome to the AI research bites. This series of short and informative talks showcases cutting-edge research work from ServiceNow AI Research team. The AI Research Bites are open to all, especially those interested in keeping up with the fast-paced AI research community.
This session features Alexandre Piché's recent work on reducing hallucinations with the ReSearch algorithm: an approach based on iterative search and self-evaluation to produce more factual generations. This work presented at SeT LLM @ ICLR 2024 leverages synthetic data to teach LLMs to abstain instead of hallucinating. To showcase the effectiveness of the ReSearch algorithm, we analyze an LLM’s ability to write biographies.
Interested in collaborating with Alex to work on this research? Feel free to reach out to his email address (see the last slide)!
Paper: https://arxiv.org/abs/2405.13022
ServiceNow AI Research team: https://www.servicenow.com/research/