February 28 – March 6: 4 frontier papers out of 1,925 read.
Papers rated as moving the field
How we choose“Results show a 63% aggregate attack success rate, sustained multi-hop propagation, and stark divergences in model security postures, highlighting that while execution-level filtering effectively mitigates dormant payloads, skill supply chains remain universally vulnerable.”
The authors, in the abstract
Self-Propagating Attacks Across LLM Agent Ecosystems
“We show this pipeline to be effective in jailbreaking OpenAI's GPT-OSS-Safeguard-20B and in prompt injections against Meta-SecAlign-70B, an adversarially robust model.”
Feb 28 – Mar 6: 58 cleared, 4 frontier
The week’s papers have not been sorted into fields yet.
Papers this week that name the model
Papers this week with an author at the lab
54 papers