English

From Actions to Understanding: Conformal Interpretability of Temporal Concepts in LLM Agents

Artificial Intelligence 2026-04-23 v1 Computation and Language Emerging Technologies Multiagent Systems Robotics

Abstract

Large Language Models (LLMs) are increasingly deployed as autonomous agents capable of reasoning, planning, and acting within interactive environments. Despite their growing capability to perform multi-step reasoning and decision-making tasks, internal mechanisms guiding their sequential behavior remain opaque. This paper presents a framework for interpreting the temporal evolution of concepts in LLM agents through a step-wise conformal lens. We introduce the conformal interpretability framework for temporal tasks, which combines step-wise reward modeling with conformal prediction to statistically label model's internal representation at each step as successful or failing. Linear probes are then trained on these representations to identify directions of temporal concepts - latent directions in the model's activation space that correspond to consistent notions of success, failure or reasoning drift. Experimental results on two simulated interactive environments, namely ScienceWorld and AlfWorld, demonstrate that these temporal concepts are linearly separable, revealing interpretable structures aligned with task success. We further show preliminary results on improving an LLM agent's performance by leveraging the proposed framework for steering the identified successful directions inside the model. The proposed approach, thus, offers a principled method for early failure detection as well as intervention in LLM-based agents, paving the path towards trustworthy autonomous language models in complex interactive settings.

Keywords

Cite

@article{arxiv.2604.19775,
  title  = {From Actions to Understanding: Conformal Interpretability of Temporal Concepts in LLM Agents},
  author = {Trilok Padhi and Ramneet Kaur and Krishiv Agarwal and Adam D. Cobb and Daniel Elenius and Manoj Acharya and Colin Samplawski and Alexander M. Berenbeim and Nathaniel D. Bastian and Susmit Jha and Anirban Roy},
  journal= {arXiv preprint arXiv:2604.19775},
  year   = {2026}
}

Comments

12 pages, 3 figures