Efficient Safety Benchmarking via Item Response Theory

arXiv:2606.20626 · cs.CY, cs.AI · Submitted 2026-05-26 · Read on arXiv

cs.CY, cs.AI

Submitted: 2026-05-26

Updated: 2026-09-25

Code: https://github.com/nd-ball/py-irt

Terminology

• Benchmark — 140× in this paper · explained in 2 episodes
• Benchmarking — 41× in this paper · explained in 3 episodes
• Item Response Theory — 27× in this paper
• Adaptive Selection — 11× in this paper · explained in 1 episode
• Fisher Information — 11× in this paper · explained in 3 episodes
• Computerized Adaptive Testing — 7× in this paper
• Accuracy — 6× in this paper · explained in 1 episode
• Robustness — 5× in this paper · explained in 5 episodes
• Redundancy — 4× in this paper · explained in 2 episodes
• Adaptive Safety — 3× in this paper · explained in 1 episode
• Alignment — 3× in this paper · explained in 1 episode
• Calibration — 3× in this paper · explained in 7 episodes
• Fidelity — 3× in this paper · explained in 2 episodes
• Adaptation — 2× in this paper · explained in 1 episode
• Information Gain — 2× in this paper · explained in 2 episodes
• International Conference on Machine Learning — 2× in this paper
• Model-Agnostic — 2× in this paper · explained in 2 episodes
• Supervised Program for Alignment Research — 2× in this paper
• Adam — 1× in this paper · explained in 2 episodes
• Adaptive Strategy — 1× in this paper · explained in 1 episode
• Adaptivity — 1× in this paper · explained in 1 episode
• Adversarial Prompting — 1× in this paper · explained in 1 episode
• Adversarial scenarios — 1× in this paper · explained in 1 episode
• Clustering — 1× in this paper · explained in 2 episodes
• LLM judge — 1× in this paper · explained in 2 episodes
• Score Function — 1× in this paper · explained in 4 episodes
• Verification — 1× in this paper · explained in 2 episodes

Sources

• JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models
• Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned
• WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs
• Stochastic Variational Inference
• The Scales of Justitia: A Comprehensive Survey on Safety Evaluation of LLMs
• HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal
• MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
• SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal
• AIR-Bench 2024: A Safety Benchmark Based on Risk Categories from Regulations and Policies
• DISCO: Diversifying Sample Condensation for Efficient Model Evaluation
• A StrongREJECT for Empty Jailbreaks
• InjecAgent: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents
• SafetyBench: Evaluating the Safety of Large Language Models
• SimpleSafetyTests: a Test Suite for Identifying Critical Safety Risks in Large Language Models
• Universal and Transferable Adversarial Attacks on Aligned Language Models
• Do-Not-Answer: A Dataset for Evaluating Safeguards in LLMs

Related papers

• Reasoning Enhances Robustness to Prompt Injection in LLM-Based Consensus
• Generative AI Purpose-built for Social and Mental Health: A Real-World Pilot
• PersonaMem-v3: Toward Omni-Platform Personal Intelligence for Holistic User Understanding, Recommendation, and Agentic Tasks
• What is an intelligent system?
• AI University: An LLM-Powered Learning Assistant for Engineering---A Finite Element Method Case Study
• Generative AI Use in Entrepreneurship: An Integrative Review and an Empowerment-Entrapment Framework