Zhi Wang

The paper surveys the use of LLMs for agentic NetOps and AIOps, arguing that operational reliability depends not on the model itself, but on robust surrounding machinery and workflow-centered evaluation.

The paper introduces Babel, an efficient black-box attack framework that systematically exploits intrinsic safety gaps in LLMs by optimizing text obfuscation sampling, achieving state-of-the-art jailbreak success rates on commercial models.

The paper introduces CORDON-MAS, a compartmentalized framework that defends Retrieval-Augmented Generation (RAG) against knowledge poisoning by enforcing strict information-flow control, significantly reducing attack success rates.

This study finds that when users do not specify a jurisdiction, the language used in the prompt strongly biases the LLM's response toward a specific national legal framework (U.S. for English, China for Mandarin Chinese), creating a risk of institutional misselection.

This paper introduces CHERRL, a controllable hacking environment for rubric-based reinforcement learning to study and mitigate reward hacking.

The paper introduces EnterpriseClawBench, an enterprise agent benchmark with 852 tasks and evaluation protocol, achieving a best configuration score of 0.663.

ASPIRE is a continual learning system that autonomously writes and refines robot control programs in a code-as-policy paradigm, discovering transferable skills and surpassing prior methods on various manipulation tasks.

The paper introduces Graph-as-Policy (GaP), a multi-agent coding harness for Variational Automation tasks that generates directed computation graphs and improves success rates and throughput through internal simulation.