arXiv AI
Aug 6
Item Response Theory for AI Safety
arXiv:2608. 05086v1 Announce Type: new Abstract: Language models differ in how safely they behave and these differences are measured by safety benchmarks.
Searching for "Harmful Refusal": A Psychometric Audit of an AI Safety Benchmark
The paper investigates whether the HELM Safety benchmark’s HarmBench dataset measures a single attribute—harmful refusal—using psychometric methods. Multidimensional item response theory indicates HarmBench does not capture a single attribute, and differential item functioning reveals that models from different developers score items differently, suggesting domain‑specific effects. The authors conclude that safety scores should only be used for attribute comparison after confirming single‑attribute validity, warning that aggregated scores can mask saturation and conflated behaviors.
Related stories
arXiv AI
Aug 12
Why Do Safety Guardrails Degrade Across Languages?
arXiv:2605. 17173v2 Announce Type: replace-cross Abstract: Large language models exhibit safety degradation in non-English languages.
arXiv AI
Jun 30
EvalSafetyGap: A Hybrid Survey and Conceptual Framework for LLM Evaluation-Safety Failures
arXiv:2606. 30219v1 Announce Type: new Abstract: LLM evaluation and AI safety face a shared measurement problem: benchmark scores, reward-model signals, and reported safety metrics can improve while the latent properties they are meant to represent remain difficult to verify.
Hugging Face Trending Papers
Jul 21
SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring
Large language models (LLMs) increasingly support science, but they can also convert hazardous scientific knowledge into actionable misuse guidance. Existing benchmarks often rely on templated queries disconnected from real-world hazards, and employ LLM-as-a-Judge paradigms without domain grounding.
arXiv AI
Jul 22
SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring
arXiv:2607. 18665v1 Announce Type: new Abstract: Large language models (LLMs) increasingly support science, but they can also convert hazardous scientific knowledge into actionable misuse guidance.
arXiv AI
Jun 11
Models That Know How Evaluations Are Designed Score Safer
arXiv:2605. 28591v2 Announce Type: replace-cross Abstract: The validity of AI safety evaluations depends on models behaving consistently across controlled and deployment settings.
arXiv AI
Oct 15
Specification-Based Testing of Large Language Models
arXiv:2610.14443v1 Announce Type: cross Abstract: Model cards, system cards, and safety reports increasingly document LLM safety behaviors, yet safety evaluations typically start from predefined benc...
arXiv AI
Jun 2
The Refusal--Compliance Tradeoff: A Large-Scale Safety Behavior Audit of Large Language Models
arXiv:2605. 05427v2 Announce Type: replace Abstract: Refusal rates are a poor proxy for LLM safety, i.
arXiv AI
Jun 30
EMPATH: A Multilingual Auditor-Judge Benchmark for Safety Evaluation of Emotional-Support Chatbots
arXiv:2606. 30256v1 Announce Type: new Abstract: Safety benchmarks often buy scalability by fixing the prompt, the language, and the turn structure.
arXiv AI
Oct 26
DeepSafe-SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring
DeepSafe‑SciHazard introduces a real‑world‑grounded benchmark for measuring scientific safety risks, comprising 2,400 hazardous and 600 oversafety questions across 12 disciplines. It defines a decomposed harm scoring system, DeHarm‑Score, that evaluates query severity, refusal behavior, and response‑level risk, further breaking down harm into executability and net‑new risk. An expert‑validation study shows DeHarm‑Score aligns 90.17% better with expert judgments than the best baseline, and the benchmark reveals that deep research agents score 32.3% higher on mean DeHarm‑Score than standard LLMs, highlighting a critical blind spot in current safety defenses.
arXiv AI
Sep 28
Efficient Safety Benchmarking via Item Response Theory
The paper demonstrates that Item Response Theory (IRT) can uncover meaningful structure in safety benchmarks for language models, allowing adaptive item selection to approximate full benchmark rankings with Spearman’s ρ > 0.90 while cutting evaluation costs by at least 80% and up to 99.9% on some suites. It also proposes a static method to extract a small, informative subset of items that can be reused across models, achieving 80–99.8% cost savings. These findings show that psychometric techniques can make safety evaluation more efficient without sacrificing ranking accuracy.
arXiv AI
Jul 10
Persona Cartography: Charting Language Model Personality Traits in Weight Space
arXiv:2607. 07916v1 Announce Type: new Abstract: Large language models exhibit recurring behavioural patterns -- personas -- that shape generalisation and safety, but we lack reliable tools for decomposing, measuring, and controlling them.