Capable but Unreliable: Canonical Path Deviation as a Causal Mechanism of Agent Failure in Long-Horizon Tasks

• URL: http://arxiv.org/abs/2602.19008v1
• Date: Sun, 22 Feb 2026 02:37:57 GMT
• Title: Capable but Unreliable: Canonical Path Deviation as a Causal Mechanism of Agent Failure in Long-Horizon Tasks
• Abstract summary: We argue that many reliability failures are caused by drift from a task's latent solution structure, not capability failures.<n>We establish this causally using a natural experiment that holds model capability and task difficulty fixed by construction.
• Score: 0.38991526486631006
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Why do language agents fail on tasks they are capable of solving? We argue that many such failures are reliability failures caused by stochastic drift from a task's latent solution structure, not capability failures. Every well-defined tool-use task imposes a canonical solution path (i.e., a convergent set of tool invocations shared across successful runs) and agent success depends critically on whether a trajectory stays within this path's operating envelope. We establish this causally using a natural experiment that holds model capability and task difficulty fixed by construction. We analyze trajectories from the Toolathlon benchmark: 22 frontier models each attempt 108 real-world tool-use tasks across 3 independent runs, yielding 515 model$\times$task units where the same model succeeds on some runs and fails on others due to LLM sampling stochasticity alone. Within these units, successful runs adhere significantly more closely to the canonical solution path than failed runs ($+$0.060 Jaccard, $p<0.0001$, $n=488$ units, 95% CI [+0.043, +0.077]). This result survives six robustness checks including cross-model-family leave-one-out validation. Critically, the causal mechanism is gradual and self-reinforcing: the adherence gap is statistically indistinguishable from zero through the first 50% of the trajectory, ruling out early-branching selection bias, and each off-canonical tool call raises the probability that the next call is also off-canonical by 22.7 percentage points ($\hatβ=+0.227$, $p<0.0001$), more than doubling the baseline rate. These findings imply that agent reliability cannot be improved by capability scaling alone, but offer a highly actionable intervention: a simple monitor that restarts the bottom tercile of runs based on mid-trajectory canonical adherence lifts success rates by $+$8.8 percentage points among intervened runs.

Related papers

• FailBench: How Reliable are VLMs at Judging Robot Task Success? [0.0]
  FailBench is a benchmark for robot failure detection comprising 2,197 manipulation attempts across 14 public sources.<n>In FailBench, 75% of failures occur naturally, and six real-world sources come from non-failure-detection datasets.
  arXiv  Detail & Related papers  (2026-09-03T09:58:55Z)
• When Not to Imitate: Boundary-Aware Skill Memory for Reliable Tool-Use LLM Agents [62.10623192614306]
  Extracting skills from past successes is critical for the efficient evolution of Large Language Model (LLM) agents.<n>We propose textbfBoundary-Aware Skill Memory (BASM), which augments each skill with explicit boundary fields.<n>BASM consistently outperforms success-distilled skill-memory baselines.
  arXiv  Detail & Related papers  (2026-08-23T10:16:06Z)
• 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)
• Tracing Agentic Failure from the Flow of Success [7.071269758592172]
  We argue that a practical failure attribution model should be lightweight and trainable without step-level supervision on failure data.<n>We propose OAT, which casts this problem as one-class learning with neural controlled differential equations.<n>With training on only 100 successful trajectories, OAT is 200--5000 $times$ faster than prompting-based baselines.
  arXiv  Detail & Related papers  (2026-07-14T13:16:14Z)
• Determinants and Limits of LLM Security-Tool Orchestration: A Study with HexStrike-AI [2.2616066835313253]
  We run 86 picoCTF challenges across seven categories and three difficulty tiers, under three tool-access regimes and three model/client configurations.<n>We apply corrections to existing tools, agent-behavior changes, and eleven new capability tools, and re-run the previously-unsuccessful trials.<n>The overall solve rate rises from 55.4% to 72.0%, and every configuration improves significantly.
  arXiv  Detail & Related papers  (2026-07-03T02:20:04Z)
• Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents [80.03824805228719]
  A major challenge in developing computer-use agents is collecting large-scale, high-quality trajectories.<n>We propose a data-centric paradigm that turns failed trajectories into agent improvements.<n>We validate this approach with the state-of-the-art OpenCUA-72B model on the OSWorld benchmark.
  arXiv  Detail & Related papers  (2026-06-30T07:44:37Z)
• Scaling Laws for Agent Harnesses via Effective Feedback Compute [53.68149869349268]
  emphEffective Feedback Compute (EFC) is a trace-level scaling coordinate that credits feedback only when it is informative, valid, non-redundant, and retained for subsequent decisions.<n>EFC-based coordinates consistently predict failure rates better than raw-compute baselines.
  arXiv  Detail & Related papers  (2026-05-28T09:45:47Z)
• Survive or Collapse: The Asymmetric Roles of Data Gating and Reward Grounding in Self-Play RL [76.45061154544568]
  Self-play reinforcement learning trains language models on their own generated tasks, co-evolving a proposer and solver without human labels.<n>Recent systems report strong reasoning gains, but collapse and instability are widely observed and poorly understood.<n>We argue instead that self-play stability is governed by two distinct levers: a data-level gate that decides which proposer-generated tasks enter the training pool, and the reward signal that updates the policy on tasks already admitted.
  arXiv  Detail & Related papers  (2026-05-21T09:19:23Z)
• Distributional Energy-Based Models for Uncertainty-Aware Structured LLM Reasoning [40.342912574072024]
  Large Language Models produce structured outputs such as travel plans or code solutions.<n>Individual reasoning steps may appear correct while the output as a whole violates budgets, fails test cases, or contradicts earlier deductions.<n>We propose a learned quality scorer with deterministic analytical constraint penalties for verifying structured LLM outputs.
  arXiv  Detail & Related papers  (2026-05-15T17:08:27Z)
• ComplexMCP: Evaluation of LLM Agents in Dynamic, Interdependent, and Large-Scale Tool Sandbox [61.862814740220806]
  $textbfComplexMCP$ is a benchmark designed to evaluate agents in rigorous conditions.<n>Built on the Model Context Protocol (MCP), $textbfComplexMCP$ provides over 300 meticulously tested tools.
  arXiv  Detail & Related papers  (2026-05-11T16:20:51Z)
• Information Fidelity in Tool-Using LLM Agents: A Martingale Analysis of the Model Context Protocol [69.11739400975445]
  We introduce the first theoretical framework for analyzing error accumulation in Model Context Protocol (MCP) agents.<n>We show that cumulative distortion exhibits linear growth and high-probability deviations bounded by $O(sqrtT)$.<n>Key findings include: semantic weighting reduces distortion by 80%, and periodic re-grounding approximately every 9 steps suffices for error control.
  arXiv  Detail & Related papers  (2026-02-10T21:08:53Z)
• Towards a Science of Scaling Agent Systems [79.64446272302287]
  We formalize a definition for agent evaluation and characterize scaling laws as the interplay between agent quantity, coordination structure, modelic, and task properties.<n>We derive a predictive model using coordination metrics, that cross-validated R2=0, enabling prediction on unseen task domains.<n>We identify three effects: (1) a tool-coordination trade-off: under fixed computational budgets, tool-heavy tasks suffer disproportionately from multi-agent overhead, and (2) a capability saturation: coordination yields diminishing or negative returns once single-agent baselines exceed 45%.
  arXiv  Detail & Related papers  (2025-12-09T06:52:21Z)
• DoVer: Intervention-Driven Auto Debugging for LLM Multi-Agent Systems [48.971606069204825]
  DoVer is an intervention-driven debug framework for large language model (LLM)-based multi-agent systems.<n>It augments hypothesis generation with active verification through targeted interventions.<n>DoVer flips 18-28% of failed trials into successes, achieves up to 16% milestone progress, and validates or refutes 30-60% of failure hypotheses.
  arXiv  Detail & Related papers  (2025-12-07T09:23:48Z)
• The 4/$δ$ Bound: Designing Predictable LLM-Verifier Systems for Formal Method Guarantee [5.345468714252351]
  This work bridges the gap by developing an LLM-Verifier Convergence Theorem.<n>We model the interaction between the LLM and the verifier as a discrete-time Markov Chain.<n>We stress-tested this prediction in an extensive empirical campaign comprising more than 90,000 trials.
  arXiv  Detail & Related papers  (2025-11-30T22:19:09Z)
• Eigen-1: Adaptive Multi-Agent Refinement with Monitor-Based RAG for Scientific Reasoning [53.45095336430027]
  We develop a unified framework that combines implicit retrieval and structured collaboration.<n>On Humanity's Last Exam (HLE) Bio/Chem Gold, our framework achieves 48.3% accuracy.<n>Results on SuperGPQA and TRQA confirm robustness across domains.
  arXiv  Detail & Related papers  (2025-09-25T14:05:55Z)
• Abduct, Act, Predict: Scaffolding Causal Inference for Automated Failure Attribution in Multi-Agent Systems [20.846301581161978]
  Failure attribution in multi-agent systems is a critical yet unsolved challenge.<n>Current methods treat this as a pattern recognition task over long conversation logs.<n>A2P Scaffolding transforms failure attribution from pattern recognition into a structured causal inference task.
  arXiv  Detail & Related papers  (2025-09-12T16:51:15Z)
• Model Discovery and Graph Simulation: A Lightweight Gateway to Chaos Engineering [0.0]
  Chaos engineering reveals resilience risks but is expensive and operationally risky to run broadly and often.<n>We claim that a simple connectivity-only topological model can provide fast, low-risk availability estimates under fail-stop faults.
  arXiv  Detail & Related papers  (2025-06-12T10:59:28Z)
• Runaway is Ashamed, But Helpful: On the Early-Exit Behavior of Large Language Model-based Agents in Embodied Environments [54.67512489842682]
  Large language models (LLMs) have demonstrated strong planning and decision-making capabilities in complex embodied environments.<n>We take a first step toward exploring the early-exit behavior for LLM-based agents.
  arXiv  Detail & Related papers  (2025-05-23T08:23:36Z)
• SURE: A Visualized Failure Indexing Approach using Program Memory Spectrum [2.4151044161696587]
  We propose SURE, a viSUalized failuRe indExing approach using the program memory spectrum. We first collect the run-time memory information at preset breakpoints during the execution of failed test cases. Any pair of PMS images that serve as proxies for two failures is fed to a trained Siamese convolutional neural network.
  arXiv  Detail & Related papers  (2023-10-19T02:04:35Z)
• Self-Evaluation Guided Beam Search for Reasoning [61.523627290397556]
  We introduce a stepwise self-evaluation mechanism to guide and calibrate the reasoning process of Large Language Model (LLM) We propose a decoding algorithm integrating the self-evaluation guidance via beam search. Our approach surpasses the corresponding Codex-backboned baselines in few-shot accuracy by $6.34%$, $9.56%$, and $5.46%$ on the GSM8K, AQuA, and StrategyQA.
  arXiv  Detail & Related papers  (2023-05-01T02:37:59Z)
• Kidney Exchange with Inhomogeneous Edge Existence Uncertainty [33.17472228570093]
  We aim to maximize a matched cycle and chain packing problem, where we aim to identify structures in a directed graph to the edge of failure. Our approaches on data from the United for Sharing (SUNO) provides better performance with the same weights as as an SAA-based method.
  arXiv  Detail & Related papers  (2020-07-07T04:08:39Z)

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.