AgentEscapeBench: Evaluating Out-of-Domain Tool-Grounded Reasoning in LLM Agents

• URL: http://arxiv.org/abs/2605.07926v1
• Date: Fri, 08 May 2026 15:59:27 GMT
• Title: AgentEscapeBench: Evaluating Out-of-Domain Tool-Grounded Reasoning in LLM Agents
• Abstract summary: AgentEscapeBench tests whether agents can infer, execute, and revise novel tool-use procedures under explicit long-range dependency constraints.<n>Experiments with sixteen LLM agents and human participants show that performance drops sharply as dependency depth increases.<n>Trajectory analysis attributes model failures mainly to breakdowns in long-range state tracking, clue adherence, and intermediate-result propagation.
• Score: 35.967227261886435
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: As LLM-based agents increasingly rely on external tools, it is important to evaluate their ability to sustain tool-grounded reasoning beyond familiar workflows and short-range interactions. We introduce AgentEscapeBench, an escape-room-style benchmark that tests whether agents can infer, execute, and revise novel tool-use procedures under explicit long-range dependency constraints. Each task defines a directed acyclic dependency graph over tools and items, requiring agents to invoke real external functions, track hidden state revealed incrementally, propagate intermediate results, and submit a deterministically verifiable final answer. AgentEscapeBench includes 270 instances across five difficulty tiers and supports fully automated evaluation. Experiments with sixteen LLM agents and human participants show that performance drops sharply as dependency depth increases: humans decline from 98.3% success at difficulty-5 to 80.0% at difficulty-25, while the best model drops from 90.0% to 60.0%. Trajectory analysis attributes model failures mainly to breakdowns in long-range state tracking, clue adherence, and intermediate-result propagation. These findings suggest that current agents can often handle local tool use but still struggle with deep contextual dependencies. We hope AgentEscapeBench can serve as a diagnostic testbed for measuring current agent capabilities and informing future training efforts toward more robust general-purpose reasoning, action, and adaptation.

Related papers

• From Sequence to Structure: Relational Uncertainty Propagation for LLM Agents [75.69180947909148]
  We propose a trajectory-level UQ framework for large language model (LLM) agents.<n>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.<n>We evaluate RUPA on representative agent benchmarks, including $$-2, Terminal-Bench-2, and GAIA, using 6 open-source LLMs spanning multiple model families.
  arXiv  Detail & Related papers  (2026-08-17T01:40:14Z)
• ScrambleToolBench: Agents Search Exhaustively Even When Their Own Map Points to the Next Step [74.83535434786145]
  We introduce ScrambleToolBench, an interactive terminal benchmark designed to isolate behavioral reasoning.<n>The benchmark requires agents to uncover hidden tool behaviors entirely through trial-and-error interaction.<n>Our evaluation of state-of-the-art language models reveals that successful initial discovery does not translate into robust adaptation.
  arXiv  Detail & Related papers  (2026-08-03T15:07:46Z)
• When Agents Do Not Stop: Uncovering Infinite Agentic Loops in LLM Agents [7.986500985812644]
  We call this problem Infinite Agentic Loops (IALs)<n>IALs arise from the interaction between agent logic, framework semantics, runtime observations, and termination mechanisms.<n>We propose IAL-Scan, a static analysis tool for detecting IAL failures in real-world LLM agent projects.
  arXiv  Detail & Related papers  (2026-07-02T03:14:27Z)
• Beyond the Black Box: Interpretability of Agentic AI Tool Use [0.0]
  We introduce a mechanistic-interpretability toolkit built on Sparse Autoencoders and linear probes.<n>The framework reads model states before each action and infers both whether a tool is needed and how consequential the next tool action is likely to be.<n>We train the probes on multi-step trajectories from the NVIDIA Nemotron function-calling dataset and apply the same workflow to GPT-OSS 20B and Gemma 3 27B models.
  arXiv  Detail & Related papers  (2026-05-07T19:47:30Z)
• AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation [71.49152943451328]
  We introduce a benchmark AJ-Bench to evaluate Agent-as-a-Judge across three domains-search, data systems, and graphical user interfaces-comprising 155 tasks and 516 annotated trajectories.<n>Experiments demonstrate consistent performance gains over LLM-as-a-Judge baselines, while also revealing substantial open challenges in agent-based verification.
  arXiv  Detail & Related papers  (2026-04-20T13:23:38Z)
• AgentProcessBench: Diagnosing Step-Level Process Quality in Tool-Using Agents [50.481033105867205]
  We introduce AgentProcessBench, the first benchmark dedicated to evaluating step-level effectiveness in realistic, tool-augmented trajectories.<n>The benchmark comprises 1,000 diverse trajectories and 8,509 human-labeled step annotations with 89.1% inter-annotator agreement.<n>It features a ternary labeling scheme to capture exploration and an error propagation rule to reduce labeling ambiguity.
  arXiv  Detail & Related papers  (2026-03-15T16:13:58Z)
• AgentRaft: Automated Detection of Data Over-Exposure in LLM Agents [40.88941407116349]
  We present AgentRaft, the first automated framework for detecting Data Over-Exposure (DOE) risks in Large Language Model (LLM) agents.<n>We evaluate AgentRaft on a testing environment of 6,675 real-world agent tools.
  arXiv  Detail & Related papers  (2026-03-08T09:40:54Z)
• FT-Dojo: Towards Autonomous LLM Fine-Tuning with Language Agents [25.60249598832918]
  FT-Dojo is an interactive environment comprising 13 tasks across 5 domains.<n>We develop FT-Agent, an autonomous system that mirrors human experts by leveraging evaluation-driven feedback.
  arXiv  Detail & Related papers  (2026-03-02T10:37:11Z)
• Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning [84.70211451226835]
  Large Language Model (LLM) Agents are constrained by a dependency on human-curated data.<n>We introduce Agent0, a fully autonomous framework that evolves high-performing agents without external data.<n>Agent0 substantially boosts reasoning capabilities, improving the Qwen3-8B-Base model by 18% on mathematical reasoning and 24% on general reasoning benchmarks.
  arXiv  Detail & Related papers  (2025-11-20T05:01:57Z)
• OmniEAR: Benchmarking Agent Reasoning in Embodied Tasks [52.87238755666243]
  We present OmniEAR, a framework for evaluating how language models reason about physical interactions, tool usage, and multi-agent coordination in embodied tasks.<n>We model continuous physical properties and complex spatial relationships across 1,500 scenarios spanning household and industrial domains.<n>Our systematic evaluation reveals severe performance degradation when models must reason from constraints.
  arXiv  Detail & Related papers  (2025-08-07T17:54:15Z)
• More Vulnerable than You Think: On the Stability of Tool-Integrated LLM Agents [24.84276066855418]
  This study investigates whether agents are vulnerable to errors throughout the entire tool invocation process.<n>We observe that agents are highly susceptible to errors at each stage and agents based on open-source models are more vulnerable than those based on proprietary models.
  arXiv  Detail & Related papers  (2025-06-27T07:13:29Z)
• Adaptive Tool Use in Large Language Models with Meta-Cognition Trigger [49.81945268343162]
  We propose MeCo, an adaptive decision-making strategy for external tool use.<n>MeCo quantifies metacognitive scores by capturing high-level cognitive signals in the representation space.<n>MeCo is fine-tuning-free and incurs minimal cost.
  arXiv  Detail & Related papers  (2025-02-18T15:45:01Z)

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.