@article{removingsandbagginginllmsbytrainingwithw,
  title = {Removing Sandbagging in LLMs by Training with Weak Supervision},
  author = {Emil Ryd and Henning Bartsch and Julian Stastny and Joe Benton and Vivek Hebbar},
  year = {2026},
  eprint = {2604.22082},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2604.22082},
}