DelegationBench: Measuring When AI Agents Should Ask Before Acting

arXiv:2610.05532 · cs.AI, cs.CL, cs.HC, cs.LG · Submitted 2026-10-04 · Read on arXiv

cs.AI, cs.CL, cs.HC, cs.LG

Submitted: 2026-10-04

Updated: 2026-10-04

Code: https://github.com/PieLord757/delegation-bench

Terminology

• AI Agents — 1× in this paper · explained in 2 episodes

Sources

• Janus: a Playground for User-Involved Agentic Permission Management
• Mind the GAP: Text Safety Does Not Transfer to Tool-Call Safety in LLM Agents
• Says Block, Still Acts: Why LLM Safety Judgments Fail to Govern Action in LLM Agents
• AgentRedBench: Dynamic Redteaming and Integration-Aware Defense for LLM Agents over SaaS Integrations
• Safety Under Scaffolding: How Evaluation Conditions Shape Measured Safety
• Coding Agents Are Guessing: Measuring Action-Boundary Violations in Underspecified DevOps Instructions
• Taxonomy and Consistency Analysis of Safety Benchmarks for AI Agents
• AgentBench: Evaluating LLMs as Agents
• AgentAbstain: Do LLM Agents Know When Not to Act?
• AgentAtlas: Beyond Outcome Leaderboards for LLM Agents
• How Agents Ask for Permission: User Permissions for AI Agents, from Interfaces to Enforcement
• Assistant or Actor? Student Trust, Control, and Delegation Regret When Using a General-Purpose AI Agent
• Overeager Coding Agents: Measuring Out-of-Scope Actions on Benign Tasks
• SteerBench-Work: A Benchmark for Agent Steering at Action Boundaries
• LM Agents May Fail to Act on Their Own Risk Knowledge
• The Authorization-Execution Gap Is a Major Safety and Security Problem in Open-World Agents
• Do Coding Agents Understand Least-Privilege Authorization?
• AgentBoundary: Counterfactual Evaluation of Safety in Tool-Using LLM Agents
• Intent-Governed Tool Authorization for AI Agents
• Capability Gates Are Not Authorization: Confused-Deputy Failures in LLM Agent Frameworks

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems