[ICML 2026] AutoControl Arena: Frontier AI Risk Auto-Discovery Platform
-
Updated
May 2, 2026 - Python
#
safety-alignment
Here are 18 public repositories matching this topic...
[CVPR'26] SafeGRPO: Self-Rewarded Multimodal Safety Alignment via Rule-Governed Policy Optimization
-
Updated
Feb 19, 2026 - Python
Ferret: Faster and Effective Automated Red Teaming with Reward-Based Scoring Technique
-
Updated
Aug 22, 2024 - Python
[ICLR26] Safety Mirage: How Spurious Correlations Undermine VLM Safety Fine-tuning
machine-unlearning trustworthy-machine-learning vision-language-model multimodal-large-language-models safety-alignment
-
Updated
Apr 16, 2026 - Python
[EMNLP 2024] Holistic Automated Red Teaming for Large Language Models through Top-Down Test Case Generation and Multi-turn Interaction
-
Updated
Nov 9, 2024 - Python
[ICML 2025] Vulnerability-Aware Alignment: Mitigating Uneven Forgetting in Harmful Fine-Tuning
-
Updated
Jun 10, 2025 - Python
AgentAlign: Navigating Safety Alignment in the Shift from Informative to Agentic Large Language Models
-
Updated
Jun 2, 2025 - Python
Official Implementation of "PSAlign: Personalized Safety Alignment for Text-to-Image Diffusion Models"
-
Updated
Jan 17, 2026 - Python
This is the official implementation (code, data) of the paper "MOSSBench: Is Your Multimodal Language Model Oversensitive to Safe Queries?""
-
Updated
Nov 16, 2024 - JavaScript
SEALGuard: Safeguarding the Multilingual Conversations in Southeast Asian Languages for LLM Software Systems
-
Updated
Jul 12, 2025 - Python
Code for the paper: "Co-Evolutionary Multi-Modal Alignment via Structured Adversarial Evolution".
-
Updated
Mar 1, 2026 - Python
Adversarial testing of LLMs on constraint satisfaction deadlocks
reinforcement-learning gemini grok claude hallucination prompt-engineering chain-of-thought chatgpt rlhf qwen llm-evaluation sycophancy deepseek safety-alignment ai-red-teaming kimi-k2 adversarial-testing
-
Updated
Jan 27, 2026
[ICML 2026] Contrastive Reasoning Alignment: Reinforcement Learning from Hidden Representations
reinforcement-learning jailbreak defence representation-learning contrastive-learning safety-alignment grpo
-
Updated
Mar 29, 2026 - Python
A PID-Lagrangian based Reinforcement Unlearning framework
-
Updated
May 1, 2026 - Python
Risk-Aware Introspective RAG (RAI-RAG) is a safety-aligned RAG framework integrating introspective reasoning, risk-aware retrieval gating, and secure evidence filtering to build trustworthy, robust, and secure LLM and agentic AI systems.
trustworthy-ai safety-alignment retrieval-augmented-generation-rag agentic-ai-security llm-security-compliance-prompt-injection ai-safety-research risk-aware-ai introspective-reasoning
-
Updated
Mar 7, 2026 - Python
Adversarial Distillation at the Frontier: Technical Mechanisms, Documented Campaigns, and a Policy Framework for Protecting American AI Capabilities — arXiv cs.AI preprint
arxiv research-paper knowledge-distillation ai-security ai-policy deepseek adversarial-distillation safety-alignment export-controls frontier-ai nstm4 us-china-ai
-
Updated
Apr 26, 2026 - TeX
Mechanistic study of the refusal direction across base, instruction-tuned, and reasoning-distilled Qwen2.5-1.5B variants: extraction, ablation, transplant, and phase-aware analysis.
jailbreak language-models ai-safety llm mechanistic-interpretability transformerlens qwen safety-alignment reasoning-language-models reasoning-models deepseek-r1 refusal-ablation refusal-direction
-
Updated
May 8, 2026 - Python
Quantifying the "Safety Half-Life" of LLMs: A framework to measure how safety alignment degrades and susceptibility to jailbreaks increases as context length grows
benchmarking machine-learning jailbreak evaluation transformers ai-safety model-drift llm context-window safety-alignment
-
Updated
Feb 2, 2026 - Python
Improve this page
Add a description, image, and links to the safety-alignment topic page so that developers can more easily learn about it.
Add this topic to your repo
To associate your repository with the safety-alignment topic, visit your repo's landing page and select "manage topics."