news
| May 29, 2026 | Our new paper Triaging Threats to Specialized Guardrails is now available on arXiv. In this work, we introduce GuardZoo, a unified human-annotated benchmark for safety guardrails, and RouteGuard, a router-expert framework for threat-specific LLM safety detection. |
|---|---|
| Apr 07, 2026 | Excited to share that our paper RedCoder: Automated Multi-Turn Red Teaming for Code LLMs has been accepted to ACL 2026 Main! RedCoder introduces an automated multi-turn red-teaming agent for stress-testing the security boundaries of Code LLMs. See you all at SD! |
| Mar 18, 2026 | Our paper DebugLM: Learning Traceable Training Data Provenance for LLMs is now available on arXiv. DebugLM enables LLMs to trace model behaviors back to their responsible training data sources and supports targeted test-time remediation without retraining. |