NPO enables stable unlearning of 50%+ training data in LLMs on TOFU by making collapse exponentially slower than gradient ascent, preserving sensible outputs where prior methods fail.
Artificial intelligence and biological misuse: Differentiating risks of language models and biological design tools [Internet]
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
representative citing papers
A trace-based benchmark of 30 security tasks finds that less-restricted LLM derivatives outperform stock safety-aligned models on some agent tasks for Gemma but not Qwen or Llama, with similar patterns on non-security controls.
SLIP enables self-jailbreaking of aligned LLMs via lexical insertion in breadth-first tree search, reaching 94.7% average ASR on AdvBench and HarmBench across eleven models with ~7.9 calls.
LLM unlearning is reframed as inadvertently installing backdoor triggers on forget-tokens; Random Noise Augmentation is introduced as a defense that improves robustness with theoretical guarantees.
Agentic biological AI systems like Biomni and K-Dense assist with dual-use tasks blocked by safeguards and gain performance uplift on WMDP proxies; BioVeil MATRIX is introduced as a 10-category taxonomy with 22 techniques to categorize and red-team AI-enabled biosecurity risks.
AI model evaluations for biological capabilities should prioritize high-consequence risks like pandemics, informed by life sciences dual-use experience, and occur prior to deployment to enable biosafety measures.
citing papers explorer
-
Negative Preference Optimization: From Catastrophic Collapse to Effective Unlearning
NPO enables stable unlearning of 50%+ training data in LLMs on TOFU by making collapse exponentially slower than gradient ascent, preserving sensible outputs where prior methods fail.
-
Measuring Safety Alignment Effects in Autonomous Security Agents
A trace-based benchmark of 30 security tasks finds that less-restricted LLM derivatives outperform stock safety-aligned models on some agent tasks for Gemma but not Qwen or Llama, with similar patterns on non-security controls.
-
Break Me If You Can: Self-Jailbreaking of Aligned LLMs via Lexical Insertion Prompting
SLIP enables self-jailbreaking of aligned LLMs via lexical insertion in breadth-first tree search, reaching 94.7% average ASR on AdvBench and HarmBench across eleven models with ~7.9 calls.
-
Improving LLM Unlearning Robustness via Random Perturbations
LLM unlearning is reframed as inadvertently installing backdoor triggers on forget-tokens; Random Noise Augmentation is introduced as a defense that improves robustness with theoretical guarantees.
-
BioVeil MATRIX: Uncovering and categorizing vulnerabilities of agentic biological AI scientists
Agentic biological AI systems like Biomni and K-Dense assist with dual-use tasks blocked by safeguards and gain performance uplift on WMDP proxies; BioVeil MATRIX is introduced as a 10-category taxonomy with 22 techniques to categorize and red-team AI-enabled biosecurity risks.
-
Prioritizing High-Consequence Biological Capabilities in Evaluations of Artificial Intelligence Models
AI model evaluations for biological capabilities should prioritize high-consequence risks like pandemics, informed by life sciences dual-use experience, and occur prior to deployment to enable biosafety measures.