Can Agents Generalize to the Open World? Unveiling the Fragility of Static Training in Tool Use

• URL: http://arxiv.org/abs/2607.01084v1
• Date: Wed, 01 Jul 2026 15:40:25 GMT
• Title: Can Agents Generalize to the Open World? Unveiling the Fragility of Static Training in Tool Use
• Abstract summary: We formalize OpenAgent, a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions.<n>Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation.<n>We propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments.
• Score: 14.813603277906623
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While Large Language Model (LLM) agents demonstrate proficiency in static benchmarks, their deployment in real-world scenarios is hindered by the dynamic nature of user queries, tool sets, and interaction dynamics. To address this generalization gap, we formalize OpenAgent (Tool-Use Agent in Open-World), a problem setting characterized by distributional shifts across query, action, observation, and domain dimensions. To systematically diagnose its impact, we construct a controlled sandbox environment where we define fine-grained environmental shifts across a four-tier hierarchy, Perception, Interaction, Reasoning, and Internalization, and conduct a comprehensive series of experiments. Our analysis yields a series of key insights, demonstrating that agents trained via both Supervised Fine-Tuning(SFT) and Reinforcement Learning suffer from varying degrees of performance degradation when confronting open environmental shifts. Building on these insights, we propose Perturbation-Augmented Fine-Tuning, a disturbance-based intervention strategy for SFT that lays the foundation for enhancing agent robustness and utility in realistic environments. Our code will be released at: https://github. com/LAMDA-NeSy/OpenAgent.

Related papers

• Are Android GUI Agents Robust Against Runtime Anomalies? AnTrap: Evaluating Agents in Dynamic Adversarial Environments [19.009627117174947]
  We introduce AnTrap, a comprehensive benchmark that injects dynamic perturbations into agent execution trajectories.<n>We propose a taxonomy organizing real-world anomalies into four layers (State, Thinking, Action and Round) with ten fine-grained subcategories.<n>We also develop a construction pipeline that preserves task solvability while introducing realistic adversarial conditions.
  arXiv  Detail & Related papers  (2026-08-25T05:57:25Z)
• EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning [57.312638079204255]
  Training large language model agents for long-horizon tool use typically relies on interactions with real or synthesized executable environments.<n>We introduce EnvACE, an agentic reinforcement learning method that replaces external environment interaction during training with world rehearsal.
  arXiv  Detail & Related papers  (2026-08-06T15:54:36Z)
• Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments [68.02342920827135]
  We propose NoisyAgent, an agentic training framework that explicitly incorporates environmental imperfections into the agent learning process.<n>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.<n>Extensive experiments demonstrate that our approach consistently improves agent robustness under noisy and dynamic environments.
  arXiv  Detail & Related papers  (2026-05-26T16:02:00Z)
• AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions [78.49000936275773]
  We introduce AgentHijack, a benchmark designed to evaluate the robustness of computer-use agents under common corruptions.<n>We evaluate a variety of desktop tasks that utilize MLLM-based agents and discover that even minor instances of corruption can result in substantial performance degradation.<n>We propose AgentHijack-Agent, a framework that integrates an action generator with enhanced grounding capabilities and an onlooker responsible for behavior summarization and environment checking.
  arXiv  Detail & Related papers  (2026-05-25T11:09:22Z)
• IndusAgent: Reinforcing Open-Vocabulary Industrial Anomaly Detection with Agentic Tools [23.270402833986605]
  IndusAgent is a tool-augmented agentic framework for open-vocabulary anomaly detection.<n>We show that IndusAgent achieves state-of-the-art zero-shot performance among all existing methods.
  arXiv  Detail & Related papers  (2026-05-20T03:52:21Z)
• Do Enterprise Systems Need Learned World Models? The Importance of Context to Infer Dynamics [16.09809323840653]
  We show that offline-trained world models can perform well in-distribution but degrade as dynamics change, whereas discovery-based agents are more robust under shift.<n>Our findings suggest that, in enterprise environments, agents should not rely solely on fixed internalized dynamics, but should incorporate mechanisms for discovering relevant transition logic at runtime.
  arXiv  Detail & Related papers  (2026-05-12T14:22:10Z)
• AgentNoiseBench: Benchmarking Robustness of Tool-Using LLM Agents Under Noisy Condition [72.24180896265192]
  We introduce AgentNoiseBench, a framework for evaluating robustness of agentic models under noisy environments.<n>We first conduct an in-depth analysis of biases and uncertainties in real-world scenarios.<n>We then categorize environmental noise into two primary types: user-noise and tool-noise.<n>Building on this analysis, we develop an automated pipeline that injects controllable noise into existing agent-centric benchmarks.
  arXiv  Detail & Related papers  (2026-02-11T20:33:10Z)
• Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning [62.499592503950026]
  Large language model (LLM) have empowered autonomous agents to perform complex tasks that require multi-turn interactions with tools and environments.<n>We propose Agent World Model (AWM), a fully synthetic environment generation pipeline.<n>We scale to 1,000 environments covering everyday scenarios, in which agents can interact with rich toolsets.
  arXiv  Detail & Related papers  (2026-02-10T18:55:41Z)
• CuES: A Curiosity-driven and Environment-grounded Synthesis Framework for Agentic RL [35.086788669916594]
  Large language model based agents are increasingly deployed in complex, tool augmented environments.<n>Existing approaches typically assume predefined task collections, an assumption that fails in novel environments.<n>We propose CuES, a Curiosity driven and Environment grounded Synthesis framework that autonomously generates diverse, executable, and meaningful tasks.
  arXiv  Detail & Related papers  (2025-12-01T06:11:37Z)
• Grounded Test-Time Adaptation for LLM Agents [75.62784644919803]
  Large language model (LLM)-based agents struggle to generalize to novel and complex environments.<n>We propose two strategies for adapting LLM agents by leveraging environment-specific information available during deployment.
  arXiv  Detail & Related papers  (2025-11-06T22:24:35Z)
• Agent-Environment Alignment via Automated Interface Generation [10.469261669159367]
  Large language model (LLM) agents have shown impressive reasoning capabilities in interactive decision-making tasks.<n> mismatches often happen between the internal expectations of the agent regarding the influence of its issued actions and the actual state transitions in the environment.<n>We propose textbfALIGN, an interface that enhances both the static information of the environment and the step-wise observations returned to the agent.
  arXiv  Detail & Related papers  (2025-05-27T11:44:50Z)
• Generalization Across Observation Shifts in Reinforcement Learning [13.136140831757189]
  We extend the bisimulation framework to account for context dependent observation shifts. Specifically, we focus on the simulator based learning setting and use alternate observations to learn a representation space. This allows us to deploy the agent to varying observation settings during test time and generalize to unseen scenarios.
  arXiv  Detail & Related papers  (2023-06-07T16:49:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.