The papers

Most central to this idea first, not most recent.

• Research Labcs.AI

  Learned latent links can undermine safety in multi-agent systems

  CISPA Helmholtz Center for Information Security · Oct 2026

• Research Labcs.AI

  TRACE reduces multi-turn attack success while largely preserving model utility

  PRADA Lab, King Abdullah University of Science and Technology · Oct 2026

• Big Techcs.LG

  Lightweight constitutional adapters improve jailbreak defenses without adversarial training

  Anthropic Fellows Program, Independent · Sept 2026

• Chinese Techcs.AI

  Safety training works better when models ignore superficial prompt cues

  Zhejiang University, Ant Group · Sept 2026

• Chinese Techcs.AI

  Selective distillation improves safety alignment while halving rollout compute and preserving reasoning.

  Zhejiang University, Ant Group · Sept 2026

• Research Labcs.AI

  Learned latent links can undermine safety in multi-agent systems

  CISPA Helmholtz Center for Information Security · Oct 2026

• Big Techcs.AIcode

  Training on aligned data can cause misalignment in different contexts

  University of Southern California, Carnegie Mellon University · Oct 2026

• Big Techcs.CL

  Reconstructing text under safety labels produces more reliable guard models

  University of Neuchâtel, Delft University of Technology · Sept 2026

• Big Techcs.AI

  Safety instruction influence controlled via eigenvalue modulation in transformers

  Google Research, Cambridge University · Sept 2026

• Industrycs.CL

  Some harmful fine-tuning examples drive far more misalignment than others

  EleutherAI · Sept 2026

• Big Techcs.CR

  Sampled text can steer black-box models past safety safeguards

  University of Southern California, Adobe · Sept 2026

• Chinese Techcs.CV

  Inner pipeline safety checks match outer guardrails with fewer false positives

  Alibaba AAIG · Sept 2026

• Big Techcs.LG

  Reinforcement learning makes models’ admissions of failure hard to reproduce

  Stanford University, Anthropic · Sept 2026

• Big Techcs.CL

  Method to permanently erase unwanted LLM personas via weight edits

  University of Macau, Mila – Québec AI Institute · Oct 2026

• Independentcs.CR

  Kernel-level preemption could halt rogue agents before network escape

  Independent Researcher · Sept 2026