From Sequence to Structure: Relational Uncertainty Propagation for LLM Agents

2026-08-16

AI Summary

This study addresses the issue of error accumulation caused by neglected long-range dependencies in uncertainty quantification for LLM agents. We propose RUPA, a novel framework introducing a trajectory-level relational uncertainty propagation mechanism. By modeling execution history as a directed trajectory graph, RUPA employs graph propagation algorithms to capture cross-step risks and integrates behavioral and goal-alignment features to generate trajectory-level confidence scores. Experimental results on benchmarks such as τ-2 demonstrate that RUPA significantly outperforms existing methods, achieving more precise uncertainty estimation and early failure detection. Consequently, this approach effectively enhances the reliability of long-horizon agent execution by mitigating compounding errors inherent in sequential decision-making processes.

Abstract

Reliable uncertainty quantification (UQ) is essential for deploying large language model (LLM) agents in complex interactive environments. Existing UQ methods largely rely on local signals, such as token probabilities, predictive entropy, or per-step confidence, and therefore overlook the long-range dependencies through which errors accumulate across an execution trajectory. As a result, they may fail to identify agent failures whose causes originate several reasoning or interaction steps before the final answer. We propose RUPA (Relational Uncertainty Propagation for Agents), a trajectory-level UQ framework for LLM agents. RUPA represents an execution history as a directed trajectory graph in which reasoning states, tool interactions, and environment feedback are nodes connected by temporal and semantic dependency edges. It then propagates uncertainty over this graph to capture how execution risk accumulates and transfers across interaction steps. The propagated signal is combined with trajectory-level behavioral features and goal-alignment information to produce a confidence estimate for the full agent trajectory. We evaluate RUPA on representative agent benchmarks, including $τ$-2, Terminal-Bench-2, and GAIA, using 6 open-source LLMs spanning multiple model families. Experimental results show that RUPA consistently outperforms existing UQ methods by providing more accurate uncertainty estimates, enabling earlier failure detection, and improving uncertainty-guided agent execution across diverse agent tasks. These results demonstrate that explicitly modeling relational dependency is crucial to reliable UQ for long-horizon LLM agents, providing a practical foundation for trustworthy agent execution.

Problem

• Uncertainty Quantification
• LLM Agents
• Long-range Dependencies
• Error Accumulation
• Trajectory-level UQ

Innovation

• Relational Uncertainty Propagation
• Trajectory-level UQ
• Directed Trajectory Graph
• Long-range Dependencies
• LLM Agents