Should I Have Expressed a Different Intent? Counterfactual Generation for LLM-Based Autonomous Control

Amirmohammad Farzaneh, Salvatore D’Oro, Osvaldo Simeone
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:29657-29673, 2026.

Abstract

Large language model (LLM)-powered agents can translate high-level user intents into plans and actions in an environment. Yet after observing an outcome, users may wonder: What if I had phrased my intent differently? We introduce a framework that enables such counterfactual reasoning in agentic LLM-driven control scenarios, while providing formal reliability guarantees. Our approach models the closed-loop interaction between a user, an LLM-based agent, and an environment as a structural causal model (SCM), and leverages test-time scaling to generate multiple candidate counterfactual outcomes via probabilistic abduction. Through an offline calibration phase, the proposed conformal counterfactual generation (CCG) yields sets of counterfactual outcomes that are guaranteed to contain the true counterfactual outcome with high probability. We showcase the performance of CCG on a wireless network control use case, demonstrating significant advantages compared to naive re-execution baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-farzaneh26a, title = {Should I Have Expressed a Different Intent? {C}ounterfactual Generation for {LLM}-Based Autonomous Control}, author = {Farzaneh, Amirmohammad and D'Oro, Salvatore and Simeone, Osvaldo}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {29657--29673}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/farzaneh26a/farzaneh26a.pdf}, url = {https://proceedings.mlr.press/v306/farzaneh26a.html}, abstract = {Large language model (LLM)-powered agents can translate high-level user intents into plans and actions in an environment. Yet after observing an outcome, users may wonder: What if I had phrased my intent differently? We introduce a framework that enables such counterfactual reasoning in agentic LLM-driven control scenarios, while providing formal reliability guarantees. Our approach models the closed-loop interaction between a user, an LLM-based agent, and an environment as a structural causal model (SCM), and leverages test-time scaling to generate multiple candidate counterfactual outcomes via probabilistic abduction. Through an offline calibration phase, the proposed conformal counterfactual generation (CCG) yields sets of counterfactual outcomes that are guaranteed to contain the true counterfactual outcome with high probability. We showcase the performance of CCG on a wireless network control use case, demonstrating significant advantages compared to naive re-execution baselines.} }
Endnote
%0 Conference Paper %T Should I Have Expressed a Different Intent? Counterfactual Generation for LLM-Based Autonomous Control %A Amirmohammad Farzaneh %A Salvatore D’Oro %A Osvaldo Simeone %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-farzaneh26a %I PMLR %P 29657--29673 %U https://proceedings.mlr.press/v306/farzaneh26a.html %V 306 %X Large language model (LLM)-powered agents can translate high-level user intents into plans and actions in an environment. Yet after observing an outcome, users may wonder: What if I had phrased my intent differently? We introduce a framework that enables such counterfactual reasoning in agentic LLM-driven control scenarios, while providing formal reliability guarantees. Our approach models the closed-loop interaction between a user, an LLM-based agent, and an environment as a structural causal model (SCM), and leverages test-time scaling to generate multiple candidate counterfactual outcomes via probabilistic abduction. Through an offline calibration phase, the proposed conformal counterfactual generation (CCG) yields sets of counterfactual outcomes that are guaranteed to contain the true counterfactual outcome with high probability. We showcase the performance of CCG on a wireless network control use case, demonstrating significant advantages compared to naive re-execution baselines.
APA
Farzaneh, A., D’Oro, S. & Simeone, O.. (2026). Should I Have Expressed a Different Intent? Counterfactual Generation for LLM-Based Autonomous Control. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:29657-29673 Available from https://proceedings.mlr.press/v306/farzaneh26a.html.

Related Material