aiqbench analysis
This research introduces a novel policy control framework that uses conformal calibration to regulate AI agents, ensuring they adhere to safety constraints while maintaining the ability to explore and improve performance in high-stakes environments. By providing finite-sample guarantees without requiring specific model class assumptions or extensive hyperparameter tuning, this method offers a robust solution for safe deployment from day one across diverse applications like natural language processing and biomolecular engineering.
Conformal Policy Control
arXiv:2603.02196v4 Announce Type: replace Abstract: An agent must try new behaviors to explore and improve. In high-stakes environments, an agent that violates safety constraints may cause harm and must be taken offline, curtailing any future interaction. Imitating old behavior is safe, but excessive conservatism discourages exploration. How much behavior change is too much? We show how to use any safe reference policy as a probabilistic regulator for any optimized but untested policy. Conformal calibration on data from the safe policy determines how aggressively the new policy can act, while provably enforcing the user's declared risk tolerance. Unlike conservative optimization methods, we do not assume the user has identified the correct model class nor tuned any hyperparameters. Unlike previous conformal methods, our theory provides finite-sample guarantees even for non-monotonic bounded loss functions, and it introduces a new policy control setting. Our experiments on applications ranging from natural language question answering to biomolecular engineering show that safe exploration is not only possible from the first moment of deployment, but can also improve performance.
Abstract: An agent must try new behaviors to explore and improve. In high-stakes environments, an agent that violates safety constraints may cause harm and must be taken offline, curtailing any future interaction. Imitating old behavior is safe, but excessive conservatism discourages exploration. How much behavior change is too much? We show how to use any safe reference policy as a probabilistic regulator for any optimized but untested policy. Conformal calibration on data from the safe policy determines how aggressively the new policy can act, while provably enforcing the user's declared risk tolerance. Unlike conservative optimization methods, we do not assume the user has identified the correct model class nor tuned any hyperparameters. Unlike previous conformal methods, our theory provides finite-sample guarantees even for non-monotonic bounded loss functions, and it introduces a new policy control setting. Our experiments on applications ranging from natural language question answering to biomolecular engineering show that safe exploration is not only possible from the first moment of deployment, but can also improve performance.
International Conference on Machine Learning (ICML), 2026. v4: Revised CPC theory to (a) show enforced smoothness for likelihood-ratio control parameter and (b) for general control parameter, assume smoothness only of constrained policy $π_t^{(β)}$ w.r.t. $β$ (rather than of $(l_i - α)\cdotπ_t^{(β)}$ or of conformal weights)
Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Statistics Theory (math.ST); Machine Learning (stat.ML)
arXiv:2603.02196 [cs.AI]
(or arXiv:2603.02196v4 [cs.AI] for this version)
Full text under the publisher's open licence · 232 words
Related coverage
Robust Conformal CBF and CLF Controllers via Iterative Policy Updates
arXiv · Jun 16, 2026
Safety Screening for Voltage Control in Active Distribution Grids via Distributionally Robust Conformal Screening
arXiv · Sep 1, 2026
Presentation: APIs for Agents: Rethinking API Programs in the MCP Era
InfoQ · Sep 23, 2026
Conformal Policy Learning with Distribution-Free Safety Guarantees
arXiv · Sep 16, 2026