Sheng Zhong
3 papers · Latest:
Cryptography & Security
Usability as a Weapon: Attacking the Safety of LLM-Based Code Generation via Usability Requirements
This paper introduces UPAttack, demonstrating how usability requirements can force LLMs to generate insecure code, achieving up to 98.1% attack success.
2605.10133
Artificial IntelligenceResolving the bias-precision paradox with stochastic causal representation learning for personalized medicine
sMMD, a stochastic causal representation learning method, resolves the bias-precision paradox in personalized medicine, improving individualized treatment effects.
2605.05706
Cryptography & SecuritySecret Stealing Attacks on Local LLM Fine-Tuning through Supply-Chain Model Code Backdoors
A new attack method exploits backdoored model code to steal sensitive secrets from local LLM fine-tuning, bypassing current defenses.
2604.27426
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.