Research Manager Permalink
Published:
A shared research memory for you and your AI agents: claims, experiments and evidence as Markdown, with an MCP server for agents.
Posts by Tags
AI agents
Generative models
Flow Matching Policy
Published:
Flow matching and diffusion from probability paths to score matching, and how to train flow-based RL policies.
Offline RL
How Wrong Can Your Critic Be? Offline RL via the Evaluation Gap
Published:
Rethinking distribution shift in offline RL through the policy evaluation gap. Slides from my ICML 2026 talk.
Offline Reinforcement Learning
Published:
Why distribution shift makes offline RL hard, the main families of methods (ReBRAC, DICE, IQL), and practical advice.
Reinforcement learning
How Wrong Can Your Critic Be? Offline RL via the Evaluation Gap
Published:
Rethinking distribution shift in offline RL through the policy evaluation gap. Slides from my ICML 2026 talk.
Flow Matching Policy
Published:
Flow matching and diffusion from probability paths to score matching, and how to train flow-based RL policies.
Offline Reinforcement Learning
Published:
Why distribution shift makes offline RL hard, the main families of methods (ReBRAC, DICE, IQL), and practical advice.
Tools
Research Manager Permalink
Published:
A shared research memory for you and your AI agents: claims, experiments and evidence as Markdown, with an MCP server for agents.