analiza aiqbench
This research introduces aiXamine, a unified black-box platform that reveals critical trade-offs between safety, security, and privacy in large language models. The study demonstrates that optimizing for alignment can inadvertently increase over-refusal rates and that privacy risks are largely orthogonal to standard safety metrics. Furthermore, it identifies a specific vulnerability where off-policy distillation causes catastrophic robustness collapse, challenging current single-axis evaluation methods.
aiXamine: Unified Black-Box Evaluation of Cross-Dimensional Trade-offs in LLM Safety, Security, and Privacy
arXiv:2608.20554v1 Announce Type: cross Abstract: The critical failure modes in deployed large language models (LLMs) are cross-dimensional: a model can score 99.3 in safety alignment while refusing one in three benign queries, or improve across every capability metric while losing 21 points in privacy. Existing evaluation frameworks that assess safety, security, and privacy independently cannot detect these patterns. We introduce aiXamine, a unified black-box platform that evaluates LLM trustworthiness across safety, security, and privacy as interdependent properties. aiXamine orchestrates 46 tests across nine services through an automated red-teaming pipeline, producing hierarchical risk profiles, from prompt-level diagnostics to cross-service trade-off analytics, that enable reproducible comparison of proprietary and open-weight systems under identical conditions. Applying aiXamine to over 120 LLMs through more than 5,000 test runs, we conduct the largest joint safety, security, and privacy study to date and uncover three cross-dimensional phenomena invisible to single-axis evaluation. First, safety enforcement incurs a quantifiable safety tax: stronger alignment systematically increases over-refusal, forcing providers to choose between protection and utility. Second, privacy is near-orthogonal to other trustworthiness dimensions and not captured by standard alignment. Third, we identify and formally characterize distillation-induced robustness collapse: off-policy distillation without on-policy correction causes entropy collapse, catastrophically destroying robustness (56.9$\to$2.6) on the same base architecture. These findings, compounded by diminishing returns from scale and category-dependent safety behaviors, demonstrate that trustworthiness is inherently multi-dimensional: progress along one axis does not guarantee, and can actively undermine, progress along others, yet current alignment methods treat it as a single objective.
Abstract:The critical failure modes in deployed large language models (LLMs) are cross-dimensional: a model can score 99.3 in safety alignment while refusing one in three benign queries, or improve across every capability metric while losing 21 points in privacy. Existing evaluation frameworks that assess safety, security, and privacy independently cannot detect these patterns. We introduce aiXamine, a unified black-box platform that evaluates LLM trustworthiness across safety, security, and privacy as interdependent properties. aiXamine orchestrates 46 tests across nine services through an automated red-teaming pipeline, producing hierarchical risk profiles, from prompt-level diagnostics to cross-service trade-off analytics, that enable reproducible comparison of proprietary and open-weight systems under identical conditions. Applying aiXamine to over 120 LLMs through more than 5,000 test runs, we conduct the largest joint safety, security, and privacy study to date and uncover three cross-dimensional phenomena invisible to single-axis evaluation. First, safety enforcement incurs a quantifiable safety tax: stronger alignment systematically increases over-refusal, forcing providers to choose between protection and utility. Second, privacy is near-orthogonal to other trustworthiness dimensions and not captured by standard alignment. Third, we identify and formally characterize distillation-induced robustness collapse: off-policy distillation without on-policy correction causes entropy collapse, catastrophically destroying robustness (56.9$\to$2.6) on the same base architecture. These findings, compounded by diminishing returns from scale and category-dependent safety behaviors, demonstrate that trustworthiness is inherently multi-dimensional: progress along one axis does not guarantee, and can actively undermine, progress along others, yet current alignment methods treat it as a single objective.
Cryptography and Security (cs.CR); Machine Learning (cs.LG)
arXiv:2608.20554 [cs.CR]
(or arXiv:2608.20554v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2608.20554 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 341 cuvinte
Subiecte conexe
M3MAD-Bench: Multi-Dimensional Evaluation of Multi-Agent Debate Across Domains and Modalities
arXiv · 3 aug. 2026
A Unified Benchmark of Deep Learning Models for Multi-task 3D Brain Tumor Segmentation from Magnetic Resonance Imaging
arXiv · 3 aug. 2026
General Collaborative Intelligence: Architecting Cognition for Resilient Multi-Agent Ecosystems
arXiv · 22 sept. 2026
Cisco’s Compiler now plans for quantum network limits
Quantum Zeitgeist · 7 oct. 2026