How much more memorization do we find when we count near-verbatim cases?

Coming soon.

How to cite our work

@article{cooper2026nv,
  title   = {Estimating near-verbatim extraction risk in language models with
             decoding-constrained beam search},
  author  = {Cooper, A. Feder and Lemley, Mark A. and De Sa, Christopher and
             Duesterwald, Lea and Casasola, Allison and Hayes, Jamie and
             Lee, Katherine and Ho, Daniel E. and Liang, Percy},
  journal = {arXiv preprint arXiv:2603.24917},
  year    = {2026}
}