Back to feed
arXiv cs.AI·

Off-Policy Evaluation with Strategic Agents via Local Disclosure

Signal
72
Hype
15
In three linesOff-policy evaluation (OPE) method for strategic agents who modify covariates in response to policy. Authors propose revealing pre-strategic information via post-hoc explanations, construct a doubly robust estimator, and establish consistency under conditional log-normal cost sensitivity assumption.
Read source
Your take?
Reinforcement learningEvalsAlignment

Summary generated by Claude — human-verified