UQLM: Uncertainty Quantification for Language Models, is a Python package for UQ-based LLM hallucination detection

DeepTeam is a framework to red team LLMs and LLM systems.

Decrypted Generative Model safety files for Apple Intelligence containing filters

[NDSS'25 Best Technical Poster] A collection of automated evaluators for assessing jailbreak attempts.

Attack to induce LLMs within hallucinations

Official repository for the paper "ALERT: A Comprehensive Benchmark for Assessing Large Language Models’ Safety through Red Teaming"

Restore safety in fine-tuned language models through task arithmetic

DuoGuard: A Two-Player RL-Driven Framework for Multilingual LLM Guardrails

NeurIPS'24 - LLM Safety Landscape

Code and dataset for the paper: "Can Editing LLMs Inject Harm?"

Source code of "Leaky Thoughts: Large Reasoning Models Are Not Private Thinkers"

Official code for the paper "SafeKey: Amplifying Aha-Moment Insights for Safety Reasoning"

[COLM 2025] LoX: Low-Rank Extrapolation Robustifies LLM Safety Against Fine-tuning

Source code for "Opt-Out: Investigating Entity-Level Unlearning for Large Language Models via Optimal Transport", ACL 2025

Multi-agent debate framework for enhancing LLM safety through red-teaming prompts, feedback-driven learning, long-term memory, and diverse structured debate strategies.

SECA: Semantically Equivalent & Coherent Attacks for Eliciting LLM Hallucinations

[USENIX Security 2025] Mind the Inconspicuous: Revealing the Hidden Weakness in Aligned LLMs’ Ethical Boundaries

safety benchmarking