analiza aiqbench
AgentWorld introduces a critical simulation framework for evaluating agentic information retrieval by integrating Big Five personality traits into user populations to expose failure modes invisible in uniform testing. The study quantifies how personality-driven interactions reveal significant quality gaps, contextual drift, and specific vulnerabilities within tool-use environments that scripted benchmarks miss. This work advances the field of AI safety by providing robust metrics for stress-testing autonomous agents against diverse human behaviors and adversarial perturbations.
AgentWorld: Personality-Aware Reliability Evaluation for Agentic Information Retrieval
arXiv:2608.24076v1 Announce Type: new Abstract: Evaluation of agentic information retrieval remains limited to scripted interactions with uniform users, missing both natural personality diversity and adversarial brittleness. We present AgentWorld, a simulation framework combining (i)Big Five (OCEAN) personality-driven user populations with stateful tool-use environments; (ii)the pass$^k$ consistency metric with structured fault classification, partial-credit scoring, and dual-control handoff verification; (iii)score-thresholded training-data export in six fine-tuning formats; and (iv)an adversarial Risk Analyser that snapshots required-intermediate-state spines, branches Monte-Carlo rollouts under four task-aware perturbation types, and quantifies risk via $\Delta P / \Delta T$ scoring, Dempster--Shafer evidence fusion, and Shapley attack-category attribution. Three experiments demonstrate the framework: a conversational analytics agent across 10 OCEAN personas (240 evaluator judgments); a customer-support agent across 5 tasks $\times$ 4 persona variants; and adversarial stress-testing of 5 tasks revealing pre-existing trajectory brittleness ($V_{\min}=0.375$ without perturbation) and tool/infrastructure-layer attack dominance (Shapley: 46% system, 38% action). Personality variation surfaces failure modes uniform testing cannot expose---cross-domain leakage, contextual drift, a 0.27-point quality gap, and 50% vs. 100% pass-rate across personas on the same task---while the Risk Analyser quantifies trajectory-level brittleness that pass$^k$ alone cannot measure.
Abstract:Evaluation of agentic information retrieval remains limited to scripted interactions with uniform users, missing both natural personality diversity and adversarial brittleness. We present AgentWorld, a simulation framework combining (i)Big Five (OCEAN) personality-driven user populations with stateful tool-use environments; (ii)the pass$^k$ consistency metric with structured fault classification, partial-credit scoring, and dual-control handoff verification; (iii)score-thresholded training-data export in six fine-tuning formats; and (iv)an adversarial Risk Analyser that snapshots required-intermediate-state spines, branches Monte-Carlo rollouts under four task-aware perturbation types, and quantifies risk via $\Delta P / \Delta T$ scoring, Dempster--Shafer evidence fusion, and Shapley attack-category attribution. Three experiments demonstrate the framework: a conversational analytics agent across 10 OCEAN personas (240 evaluator judgments); a customer-support agent across 5 tasks $\times$ 4 persona variants; and adversarial stress-testing of 5 tasks revealing pre-existing trajectory brittleness ($V_{\min}=0.375$ without perturbation) and tool/infrastructure-layer attack dominance (Shapley: 46% system, 38% action). Personality variation surfaces failure modes uniform testing cannot expose---cross-domain leakage, contextual drift, a 0.27-point quality gap, and 50% vs. 100% pass-rate across personas on the same task---while the Risk Analyser quantifies trajectory-level brittleness that pass$^k$ alone cannot measure.
Artificial Intelligence (cs.AI)
arXiv:2608.24076 [cs.AI]
(or arXiv:2608.24076v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2608.24076 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 273 cuvinte
Subiecte conexe
Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
METR · 26 aug. 2026
PDEU-Bench: Benchmarking the Personalized Planning Lifecycle of Tool-Calling LLM Agents
arXiv · 29 sept. 2026
Natera’s intelligent appointment scheduling with Amazon Bedrock AgentCore
AWS · 26 aug. 2026
TeaRAG: A Token-Efficient Agentic Retrieval-Augmented Generation Framework
arXiv · 24 iul. 2026