Clarkson University

MLCommons Jailbreak Benchmark v1.0 introduces an end-to-end methodology that evaluates eight open-weight large language models with paired baseline and adversarial conditions using 264 seed prompts spanning eleven hazard categories and representative attacks drawn from the MLCommons Jailbreak Taxonomy, judging responses with the AILuminate Assessment Standard v1.4 and measuring robustness through the Resilience Gap; the unsafe-response rate rose from 11.08% at baseline to 18.65% under jailbreak conditions, yielding an average Resilience Gap of 7.57%, with accessible systems showing a larger mean gap and attack effectiveness varying substantially across attack categories and hazards.

MLCommons Jailbreak Benchmark v1.0 introduces an end-to-end methodology that evaluates eight open-weight large language models with paired baseline and adversarial conditions using 264 seed prompts spanning eleven hazard categories and representative attacks drawn from the MLCommons Jailbreak Taxonomy, judging responses with the AILuminate Assessment Standard v1.4 and measuring robustness through the Resilience Gap; the unsafe-response rate rose from 11.08% at baseline to 18.65% under jailbreak conditions, yielding an average Resilience Gap of 7.57%, with accessible systems showing a larger mean gap and attack effectiveness varying substantially across attack categories and hazards.

The authors propose Higher-Order-KANDy, which combines delay and derivative (jet) embeddings with derivative-ordered KANDy layers stacked into a machine learning architecture and learns a delay-to-jet map as a Kolmogorov-Arnold network, handling higher-order systems natively without a candidate library; on synthetic benchmarks it recovers equations under coarse sampling and noise, and at noise levels up to 7.5% it selects the exact active set on every seed, whereas an oracle weak-form regression admits a spurious term at that noise level, though the oracle remains the more accurate coefficient estimator.

The authors propose Higher-Order-KANDy, which combines delay and derivative (jet) embeddings with derivative-ordered KANDy layers stacked into a machine learning architecture and learns a delay-to-jet map as a Kolmogorov-Arnold network, handling higher-order systems natively without a candidate library; on synthetic benchmarks it recovers equations under coarse sampling and noise, and at noise levels up to 7.5% it selects the exact active set on every seed, whereas an oracle weak-form regression admits a spurious term at that noise level, though the oracle remains the more accurate coefficient estimator.