CIAware-Bench: Benchmarking Control Intervention Awareness Across Frontier LLMs

arXiv:2606.11063 · cs.AI, cs.LG · Submitted 2026-06-09 · Read on arXiv

cs.AI, cs.LG

Submitted: 2026-06-09

Updated: 2026-10-05

Code: https://github.com/JoachimSchaeffer/CIAware-Bench

Terminology

• Benchmark — 2× in this paper · explained in 2 episodes
• Accuracy — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 4 episodes

Sources

• Self-Generated Text Recognition: Quality Heuristics, Cross-Task Transfer, and Downstream Bias in LLM Evaluation
• Ctrl-Z: Controlling AI Agents via Resampling
• DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence
• Can you Finetune your Binoculars? Embedding Text Watermarks into the Weights of Large Language Models
• When can we trust untrusted monitoring? A safety case sketch across collusion strategies
• Games for AI Control: Models of Safety Evaluations of AI Deployment Protocols
• Evaluating Control Protocols for Untrusted AI Agents
• SHADE-Arena: Evaluating Sabotage and Monitoring in LLM Agents
• Auditing language models for hidden objectives
• Large Language Models Often Know When They Are Being Evaluated
• OpenAI GPT-5 System Card
• Attack Selection Reduces Safety in Concentrated AI Control Settings against Trusted Monitoring
• Instrumental Monitor Evasion Emerges Under Ordinary Task Pressure
• Prefill Awareness in Large Language Models

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