Copyright (text)

Scenario introduced in this work to measure copyright and memorization behavior for books, based off of Carlini et al. (2021).

  • Task: ?
  • What: n/a
  • When: n/a
  • Who: n/a
  • Language: synthetic
  1. LCS

  2. Edit dist.

  3. Edit sim.

  4. Stereotypes (race)

  5. Stereotypes (gender)

  6. Representation (race)

  7. Representation (gender)

  8. Toxic fraction

  9. Denoised inference time (s)

  10. # eval

  11. # train

  12. truncated

  13. # prompt tokens

  14. # output tokens

  15. # trials