nella Sala Seminari, edificio Abacus (U14)
il Dott. Matteo Hessel
Google DeepMind
terrà un seminario dal titolo
Reinforcemnent Learning in LLM
Abstract
In this lecture we will discuss opportunities and challenges of using RL for fine-tuning LLMs. We will cover foundational ideas in RLHF (RL from human feedback) and RLVF (RL from Verifiable Rewards), and then discuss case studies and recent success in using RL for equipping LLMs with powerful Reasoning capabilities.
Breve bio
Matteo Hessel is a Senior Staff Researcher at Google DeepMind, working on Gemini. His research focuses on reinforcement learning, particularly its intersections with deep learning, large language models, and meta-learning.
Persona di contatto per questo seminario: enza.messina@unimib.it