Item Response Theory for AI Safety

arXiv:2608.05086 · cs.AI, cs.CL · Submitted 2026-08-05 · Read on arXiv

Joshua Fonseca Rivera, Neil Shah, David Demitri Africa, Konstantinos Voudouris

Independent · Independent · UK AI Security Institute · UK AI Security Institute

cs.AI, cs.CL

Submitted: 2026-08-05

Comments: 15 pages, 9 figures, 6 tables

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 142× in this paper · explained in 2 episodes
• Calibration — 13× in this paper · explained in 9 episodes
• Item Response Theory — 13× in this paper
• Benchmarking — 5× in this paper · explained in 4 episodes
• Distillation — 3× in this paper · explained in 2 episodes
• LLM judge — 3× in this paper · explained in 2 episodes
• Quantization — 3× in this paper · explained in 15 episodes
• Failure Modes — 2× in this paper · explained in 3 episodes
• Regularization — 2× in this paper · explained in 3 episodes
• Robustness — 2× in this paper · explained in 6 episodes
• Comparative Fit Index — 1× in this paper
• Composition — 1× in this paper · explained in 2 episodes
• Evaluation Framework — 1× in this paper · explained in 3 episodes
• Fidelity — 1× in this paper · explained in 2 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Fisher Information — 1× in this paper · explained in 4 episodes
• Low-Resource Languages — 1× in this paper · explained in 4 episodes
• Perplexity — 1× in this paper · explained in 3 episodes
• Root Mean Square Error of Approximation — 1× in this paper

Sources

• Measuring what Matters: Construct Validity in Large Language Model Benchmarks
• OR-Bench: An Over-Refusal Benchmark for Large Language Models
• Adversarial Humanities Benchmark: Results on Stylistic Robustness in Frontier Model Safety
• Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned
• Investigating the Impact of Quantization Methods on the Safety and Reliability of Large Language Models
• metabench -- A Sparse Benchmark of Reasoning and Knowledge in Large Language Models
• The Silent Hyperparameter: Quantifying the Impact of Inference Backends on LLM Reproducibility
• Capabilities Ain't All You Need: Measuring Propensities in AI
• Efficient Safety Benchmarking via Item Response Theory
• AI Sandbagging: Language Models can Strategically Underperform on Evaluations
• Measuring What AI Systems Might Do: Towards A Measurement Science in AI
• Do-Not-Answer: A Dataset for Evaluating Safeguards in LLMs
• SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal
• Health-ORSC-Bench: A Benchmark for Measuring Over-Refusal and Safety Completion in Health Context
• Universal and Transferable Adversarial Attacks on Aligned 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