Perplexity

Perplexity is a quantitative evaluation metric used in natural language processing that measures the degree of uncertainty or surprise of a language model in predicting a test sequence, defined as the exponential of the average negative log-likelihood per token.

339

Publications

27K

Citations

875

Authors

271

Institutions

Publications per year

2017–2026

108

Authors

875

  1. MO
    Mari Ostendorf

    University of Washington, United States

    5 publications · 204 citations

    5

    h-index

  2. RL
    Roger Lévy

    Massachusetts Institute of Technology, United States

    5 publications · 325 citations

    5

    h-index

  3. EW
    Ethan Wilcox

    Massachusetts Institute of Technology, United States

    5 publications · 325 citations

    5

    h-index

  4. TK
    Tatsuya Kawahara

    Kyoto University, Japan

    5 publications · 187 citations

    5

    h-index

  5. RS
    Richard Schwartz

    National Institutes of Health, United States

    4 publications · 100 citations

    4

    h-index

1–5 of 875

Venues

  1. arXiv (Cornell University)

    Repository · 47 publications · 9.6K citations

    24

    h-index

  2. 6

    h-index

  3. ArXiv.org

    Repository · 4 publications · 365 citations

    4

    h-index

  4. 3

    h-index

  5. 3

    h-index

1–5