paperAugust 2026Unreviewed

EvoFlint: An Evolutionary Atlas of Multi-Turn LLM Vulnerabilities

Feitong Qiao, Liren Peng, Shiming Ren, Aishwarya Jadhav, Arghavan Bahadorinejad, Marinette Chen, Muhan Zhang, Abdulaziz Suria, Gennevi Lu, Anish Das Sarma

View resource PDF

Abstract

Frontier language models that refuse harmful single-turn prompts often comply when the same intent is reached gradually over many turns, making multi-turn attacks one of the least understood failure modes of large language models. Most automated red-teaming methods treat this as a generation problem: produce attacks that break the model. We argue it is better framed as a search problem: discover, organize, and iteratively refine a diverse archive of attack strategies, producing a structured map

Framework mappings

NIST AI Risk Management Framework

MEASUREMeasure

Suggested from the entry's categories.

Cite

@misc{qiao2026evoflint,
  title = {{EvoFlint: An Evolutionary Atlas of Multi-Turn LLM Vulnerabilities}},
  author = {Feitong Qiao and Liren Peng and Shiming Ren and Aishwarya Jadhav and Arghavan Bahadorinejad and Marinette Chen and Muhan Zhang and Abdulaziz Suria and Gennevi Lu and Anish Das Sarma},
  year = {2026},
  month = aug,
  eprint = {2609.00487},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.00487}
}