The Why Behind the Action: Unveiling Internal Drivers via Agentic Attribution

TL;DR

This paper introduces a hierarchical framework for understanding the internal reasons behind agent actions in large language model-based systems, enhancing accountability and safety by identifying key internal factors regardless of success or failure.

Contribution

The paper presents a novel general agentic attribution framework that operates hierarchically to identify internal drivers of agent actions, applicable across diverse scenarios and independent of task outcomes.

Findings

Framework reliably identifies pivotal historical events influencing actions.

Effective in diverse scenarios including tool use and bias detection.

Enhances accountability in autonomous agent systems.

Abstract

Large Language Model (LLM)-based agents are widely used in real-world applications such as customer service, web navigation, and software engineering. As these systems become more autonomous and are deployed at scale, understanding why an agent takes a particular action becomes increasingly important for accountability and governance. However, existing research predominantly focuses on \textit{failure attribution} to localize explicit errors in unsuccessful trajectories, which is insufficient for explaining \textbf{the reason behind agent behaviors}. To bridge this gap, we propose a novel framework for \textbf{general agentic attribution}, designed to identify the internal factors driving agent actions regardless of the task outcome. Our framework operates hierarchically to manage the complexity of agent interactions. Specifically, at the \textit{component level}, we employ temporal likelihood dynamics to identify critical interaction steps; then at the \textit{sentence level}, we refine this localization using perturbation-based analysis to isolate the specific textual evidence. We validate our framework across a diverse suite of agentic scenarios, including standard tool use and subtle reliability risks like memory-induced bias. Experimental results demonstrate that the proposed framework reliably pinpoints pivotal historical events and sentences behind the agent behavior, offering a critical step toward safer and more accountable agentic systems. Codes are available at https://github.com/AI45Lab/AgentDoG.