Beyond Decoding: Meta-Generation Algorithms for Large Language Models

Join a remote research talk exploring advanced meta-generation algorithms for Large Language Models, presented by Sean Welleck from Carnegie Mellon University as part of the Joint IFML/MPG Symposium at the Simons Institute. Delve into cutting-edge techniques that go beyond traditional decoding methods, examining novel approaches to enhance the generation capabilities of LLMs. Throughout the 82-minute presentation, discover innovative algorithmic strategies that are shaping the future of language model outputs and their practical applications.