Off-Policy Evaluation Estimates a New Policy From Old Logs
Off-policy evaluation uses logged actions, outcomes, and selection probabilities to estimate a policy not yet deployed. Poor coverage makes estimates unstable or impossible.
0
1
Contributors are:
Tags
AI Agent Graph Engineering
Graph Engineering for AI Agents
Related
Emergent Agent Communication Requires Interpretability Tests
In this situation—learning which explanation style helps a learner answer the next question—which choice best applies “Contextual Bandits Learn Routing Under Guardrails”?
In this situation—learning which tutoring intervention improved a later transfer task—which choice best applies “Credit Assignment Links Outcomes to Earlier Choices”?
Learn After
Related
Emergent Agent Communication Requires Interpretability Tests
In this situation—learning which explanation style helps a learner answer the next question—which choice best applies “Contextual Bandits Learn Routing Under Guardrails”?
In this situation—learning which tutoring intervention improved a later transfer task—which choice best applies “Credit Assignment Links Outcomes to Earlier Choices”?