I. Safety Alignment

• Towards Context-Invariant Safety Alignment for Large Language Models
  ICML, 2026
• From Sparse Decisions to Dense Reasoning: A Multi-attribute Trajectory Paradigm for Multimodal Moderation
  arxiv, 2026
• Mechanistic Origin of Moral Indifference in Language Models
  arxiv, 2026
• SafeVid: Toward Safety Aligned Video Large Multimodal Models
  NeurIPS, 2025
• A2RM: Adversarial-Augmented Reward Model
  Submitted to ICLR, 2026
• SafeWork-R1: Coevolving Safety and Intelligence under the AI-45∘ Law
  arxiv, 2025
• The Other Mind: How Language Models Exhibit Human Temporal Cognition
  AAAI, 2026
• Fake Alignment: Are LLMs Really Aligned Well?
  Naacl, 2024

II. Advanced Evaluation

• SafeEvalAgent: Toward Agentic and Self -Evolving Safety Evaluation of LLMs
  ACL findings, 2026
• GhostEI-Bench: Do Mobile Agents Resilience to Environmental Injection in Dynamic On-Device Environments?
  ICLR, 2026
• A Rigorous Benchmark with Multidimensional Evaluation for Deep Research Agents: From Answers to Reports
  arxiv, 2026
• Argus Inspection: Do Multimodal Large Language Models Possess the Eye of Panoptes?
  ACM MM, 2025
• LinguaSafe: A Comprehensive Multilingual Safety Benchmark for Large Language Models
  Arxiv, 2025
• Reflection-Bench: Evaluating Epistemic Agency in Large Language Models
  ICML, 2025

III. Attack and Defense

• OpenRT: An Open-Source Red Teaming Framework for Multimodal LLMs
  Arxiv, 2026
• Probing the robustness of large language models safety to latent Perturbations
  Submitted to ICLR, 2026
• Evolve the Method, Not the Prompts: Evolutionary Synthesis of Jailbreak Attacks on LLMs
  arxiv, 2026
• FreezeVLA: Action-Freezing Attacks on Vision-Language-Action Models
  Submitted to ICLR, 2026
• JailBound: Jailbreaking Internal Safety Boundaries of Vision-Language Models
  NeurIPS, 2025
• A Mousetrap: Fooling Large Reasoning Models for Jailbreak with Chain of Iterative Chaos
  ACL findings, 2025
• From Evasion to Concealment: Stealthy Knowledge Unlearning for LLMs
  ACL findings, 2025
• HoneypotNet: Backdoor Attacks Against Model Extraction
  AAAI, 2025
• StolenLoRA: Exploring LoRA Extraction Attacks via Synthetic Data
  ICCV, 2025
• IDEATOR: Jailbreaking and Benchmarking Large Vision-Language Models Using Themselves
  ICCV, 2025

IV. Interdisciplinary Research

• The Other Mind: How Language Models Exhibit Human Temporal Cognition
  AAAI, 2026
• From Pixels to Principles: A Decade of Progress and Landscape in Trustworthy Computer Vision
  Science and Engineering Ethics, 2024
• Collectivism and individualism political bias in large language models: A two-step approach
  Big Data & Society, 2025
• Chain of Risks Evaluation (CORE): A framework for safer large language models in public mental health
  Psychiatry and Clinical Neurosciences, 2025