Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments

2026-05-26

AI Summary

Large language model agents demonstrate strong performance in idealized settings but suffer significant degradation in real-world scenarios due to various sources of noise, such as ambiguous user instructions and tool execution failures. To address this challenge, this work proposes NoisyAgent, a novel framework that systematically models and explicitly incorporates two primary types of noise—user-induced and tool-related—into the training process. By integrating a progressive curriculum learning strategy that dynamically adjusts noise intensity during training, NoisyAgent enables robust agent learning under realistic conditions. Empirical results show that the proposed approach not only substantially enhances agent robustness across diverse noisy environments but also yields consistent performance gains on standard benchmarks, demonstrating that noise-aware training effectively promotes generalization capabilities.

Abstract

Recent advances in large language models (LLMs) have facilitated the widespread deployment of LLMs as interactive agents capable of reasoning, planning, and tool use. Despite strong performance on existing benchmarks, such agents often exhibit notable degradation when deployed in real-world settings, where environments are inherently stochastic and imperfect. We argue that this discrepancy arises from a fundamental mismatch between idealized training settings and real-world interaction dynamics, where current paradigms rely on carefully curated task instructions and stable, well-controlled environments. To address this gap, we propose NoisyAgent, an agentic training framework that explicitly incorporates environmental imperfections into the agent learning process. We identify two major sources of interaction noise in real-world scenarios: user noise, which captures ambiguity and variability in user interaction, and tool noise, which reflects failures and anomalies in tool execution. We introduce such perturbations into the training pipeline by modifying user interaction patterns and simulating tool execution results within the training environment. To stabilize training while encouraging agents to handle increasingly challenging imperfections, noise is applied to only a subset of rollouts and progressively increased in difficulty as the model adapts to the current noise level. Extensive experiments demonstrate that our approach consistently improves agent robustness under noisy and dynamic environments. Our analysis reveals that training under noise conditions also yields performance gains on idealized benchmarks, suggesting that controlled exposure to environmental noise promotes more generalizable reasoning and decision-making behaviors. Our findings highlight the importance of modeling interaction imperfections for bridging the gap between agent training and real-world deployment.

Problem

• robustness
• environmental noise
• LLM agents
• real-world deployment
• interaction imperfections

Innovation

• NoisyAgent
• environmental noise
• robustness
• interactive agents
• noise curriculum