arXiv AI
Jul 16
When Agents Disagree With Themselves: Behavioral Consistency as an Uncertainty Signal for LLM Agents
arXiv:2602. 11619v2 Announce Type: replace Abstract: Running the same LLM agent on identical inputs yields 2.
AgentPerturb: Evaluating LLM Agent Robustness under Controlled Interaction Perturbations
AgentPerturb introduces a systematic framework to evaluate large language model agents under controlled interaction perturbations, addressing the lack of unified evaluation methods for noisy real‑world tasks. The study quantifies robustness across frontier models, identifies task‑specific vulnerabilities, and shows that agents often fail to detect altered inputs, propagating errors and incurring higher computational costs. The authors provide a benchmark‑construction pipeline and plan to release code and data to support future robustness‑oriented training.
Related stories
arXiv AI
Jun 6
When Tools Fail: Benchmarking Dynamic Replanning and Anomaly Recovery in LLM Agents
arXiv:2606. 05806v1 Announce Type: new Abstract: Existing benchmarks evaluate Tool-Integrated Reasoning (TIR) in LLMs on idealized ''happy paths'', largely overlooking real-world tool failures.
arXiv AI
Jun 16
ARB4WM: An Adversarial Robustness Benchmark for World Models in Continuous Control
arXiv:2606. 16605v1 Announce Type: new Abstract: World models are widely used in robotic and agentic engineering control systems due to their ability to learn latent dynamics for planning and decision-making.
arXiv AI
Oct 16
LEASH: Uncertainty-Aware Weak-to-Strong Escalation and Oversight for Computer-Use Agents
LEASH is a training‑free framework that enables uncertainty‑aware escalation from a weak to a strong computer‑use agent. It uses latent uncertainty signals from the weak agent’s hidden states to decide when to hand off control, and a shadow observer to maintain oversight of the strong black‑box agent. Experiments on OSWorld show LEASH improves task success over standalone weak agents and outperforms random escalation, achieving comparable performance to Claude Opus 5 while cutting inference cost by 19.2%.
arXiv AI
Jul 2
Can Agents Generalize to the Open World? Unveiling the Fragility of Static Training in Tool Use
arXiv:2607. 01084v1 Announce Type: new 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.
arXiv AI
Sep 25
Understanding and Exploiting Initialization Anchoring Weakness in Feedback-Based Agent Planning
The paper investigates a vulnerability in feedback‑based agent planning, showing that the first round of feedback corrects a large portion of adversarial directions (46%) while subsequent rounds see a sharp decline (13% and 7%). The authors attribute this to an initialization anchoring weakness driven by plausible plan shifts, lack of counterevidence, and persistence of accepted directions. They introduce “InitAnchor”, a black‑box attack framework that exploits these factors, achieving high attack success rates across diverse tasks, architectures, and LLMs, and remaining effective against multiple defenses and real‑world agents.
arXiv Machine Learning
Sep 7
How to Speculate about Uncertainty in Agentic Coding? A Draft-Model Gate Method
The paper introduces Speculative Uncertainty (SU), a technique that infers a failure likelihood for black‑box LLM agents by evaluating their generated token sequences with a lightweight draft model, without needing internal model details. SU extracts phase‑aware features from reasoning and action spans, calibrates them against verifiable outcomes, and produces a failure‑likelihood score usable by downstream policies. Applying a pre‑execution veto gate based on SU to software‑engineering agents such as Qwen3‑Coder‑480B and Claude 3.5 Sonnet reduced execution error rates by 6‑8 percentage points and token costs by 14‑19 %, while maintaining performance on out‑of‑distribution benchmarks and across different agent models.
arXiv Computation and Language
Sep 1
CAST: Critique-Aware Supervision for Training Reliable Long-Horizon Tool-Calling Agents
CAST is a critique‑aware training framework that transforms sparse task outcomes into action‑level supervision for both critique learning and policy optimization. By analyzing agent trajectories, CAST synthesizes structured rationales that explain action validity under partial observability, enabling the creation of richer training data. Fine‑tuned Qwen3‑family models trained with CAST show significant reliability gains, outperforming GPT‑OSS‑120B by over 10% on Retail tasks and improving Telehealth performance by 9% in an out‑of‑domain setting.
arXiv Machine Learning
Oct 8
RAISED: Self-Distillation for Robustness to Prompt Injection in LLM Agents
arXiv:2610.06401v2 Announce Type: replace-cross Abstract: Tool-using language-model agents are vulnerable to indirect prompt injection because they must act on untrusted external content. Existing tr...
arXiv AI
Oct 1
SEABench: Benchmarking Endogenous Misalignment In Self-Evolving Agents
arXiv:2609.35596v2 Announce Type: replace-cross Abstract: Self-evolving LLM agents have gained prominence for their ability to improve after deployment by modifying their harness, including their con...
arXiv Machine Learning
Jun 18
Stealthy World Model Manipulation via Data Poisoning
arXiv:2606. 18697v1 Announce Type: new Abstract: Model-based learning agents use learned world models to predict future states, plan actions, and adapt to new environments.
Hugging Face Trending Papers
Sep 28
SEABench: Benchmarking Endogenous Misalignment In Self-Evolving Agents
SEABench is a benchmark designed to study endogenous misalignment in self‑evolving large language model agents. It contains 48 longitudinal task sequences across various evolution surfaces, task domains, and harm types, and includes an adaptive trajectory discovery pipeline that probes for failures while preserving task intent. Evaluations show that self‑evolution improves task completion rates but often introduces safety failures absent in non‑evolving baselines, with divergent safety behaviors reflected in agents’ chain‑of‑thought reasoning that can be monitored to mitigate unsafe actions.