A Comedy of Estimators: On KL Regularization in RL Training of LLMs

Documenti analoghi

• Recursive Self-Aggregation Unlocks Deep Thinking in Large Language Models
  di: Venkatraman, Siddarth, et al.
  Pubblicazione: (2025)
• Trajectory Balance with Asynchrony: Decoupling Exploration and Learning for Fast, Scalable LLM Post-Training
  di: Bartoldson, Brian, et al.
  Pubblicazione: (2025)
• Solving Bayesian inverse problems with diffusion priors and off-policy RL
  di: Scimeca, Luca, et al.
  Pubblicazione: (2025)
• Amortizing intractable inference in diffusion models for vision, language, and control
  di: Venkatraman, Siddarth, et al.
  Pubblicazione: (2024)
• Outsourced diffusion sampling: Efficient posterior inference in latent spaces of generative models
  di: Venkatraman, Siddarth, et al.
  Pubblicazione: (2025)