TY - JOUR
T1 - Counterfactual Planning for Generalizable Agents’ Actions
AU - Fu, Jiarun
AU - Ding, Lizhong
AU - Wei, Qiuning
AU - Guo, Yuhan
AU - Cheng, Yurong
AU - Zhang, Junyu
N1 - Publisher Copyright:
© 2026, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved.
PY - 2026
Y1 - 2026
N2 - Large language models have revolutionized agent planning by serving as the engine of heuristic guidance. However, LLM-based agents often struggle to generalize across complex environments and to adapt to stochastic feedback arising from environment–action interactions. We propose Counterfactual Planning—a method designed to improve the generalizability and adaptability of agents’ actions by inferring causal representations of environmental confounders and performing counterfactual reasoning over planned actions. We formalize the agent planning process as a structural causal model, providing a mathematical formulation for causal analysis of how environmental states influence action generation and how actions affect future state transitions. To support generalizable action planning, we introduce the State Causality Evaluator (SCE), which dynamically infers task-conditioned causal representations from complex environment states; and to enhance adaptability under stochastic feedback, we propose the What-If-Not (WIN) reward, which performs counterfactual interventions to refine actions through causal evaluation. We validate our framework in an open-world environment, where experiments demonstrate improvements in both action generalization and planning adaptability.
AB - Large language models have revolutionized agent planning by serving as the engine of heuristic guidance. However, LLM-based agents often struggle to generalize across complex environments and to adapt to stochastic feedback arising from environment–action interactions. We propose Counterfactual Planning—a method designed to improve the generalizability and adaptability of agents’ actions by inferring causal representations of environmental confounders and performing counterfactual reasoning over planned actions. We formalize the agent planning process as a structural causal model, providing a mathematical formulation for causal analysis of how environmental states influence action generation and how actions affect future state transitions. To support generalizable action planning, we introduce the State Causality Evaluator (SCE), which dynamically infers task-conditioned causal representations from complex environment states; and to enhance adaptability under stochastic feedback, we propose the What-If-Not (WIN) reward, which performs counterfactual interventions to refine actions through causal evaluation. We validate our framework in an open-world environment, where experiments demonstrate improvements in both action generalization and planning adaptability.
UR - https://www.scopus.com/pages/publications/105034247639
U2 - 10.1609/aaai.v40i35.40184
DO - 10.1609/aaai.v40i35.40184
M3 - Conference article
AN - SCOPUS:105034247639
SN - 2159-5399
VL - 40
SP - 29432
EP - 29440
JO - Proceedings of the AAAI Conference on Artificial Intelligence
JF - Proceedings of the AAAI Conference on Artificial Intelligence
IS - 35
T2 - 40th AAAI Conference on Artificial Intelligence, AAAI 2026
Y2 - 20 January 2026 through 27 January 2026
ER -