Rezultati pretraživanja
  1. 19. lip 2019.

    New paper analyzing attention in GPT-2: Some highlights: attention targets different parts of speech at different layer depths, and a simple algorithm reveals highly specialized generation patterns (e.g. acronyms from names). With .

  2. 1. stu 2018.

    You can remove most of the question's words and still get a correct answer on SQuAD

  3. 12. lip 2019.

    Check out our new paper "What Does BERT Look At? An Analysis of BERT's Attention" with ! Among other things, we show that BERT's attention corresponds surprisingly well to aspects of syntax and coreference.

  4. 1. stu 2018.

    Hupkes et al.: When the text has disfluencies - attention does the heavy lifting, decoder makes heavy use of it, encoder isn't really doing anything special other than preparing the input for the other modules to process

  5. Can we model a black-box with a black-box? -- Leila Wehbe

  6. 24. svi 2019.

    I wrote "An Analysis of Source-Side Grammatical Errors for NMT", basically using an eng->deu system to translate noisy and corrected versions of sentences from GEC corpora and analyzing the outputs. Accepted at 2019. Any feedback welcome!

  7. 1. stu 2018.

    Looks like we have at the Chit Chat tutorial.

  8. 2. ruj 2018.

    Not so soon, but now availble at

    Prikaži ovu nit
  9. 1. stu 2018.

    RNN LMs can learn some but not all island constraints (E. Wilcox, R. Levy, T. Morita & R. Futrell). Subject islands are hard. Yet unanswered Q (paraphrased M. Baroni): does gradience suggest islands aren't grammatical, but are instead likely processing?

  10. 1. stu 2018.
    Odgovor korisnicima

    Thank you! Looking forward to next year’s !

  11. Just noticed that the latest release of includes tools for model interpretability. Not sure yet how useful they are, but happy to see this given some attention in a mainstream framework.

    Prikaži ovu nit
  12. 1. stu 2018.

    It's an honor to see in person at and and get his words written on his authored book!

  13. 6. lip 2019.

    Our paper "On the Realization of Compositionality in Neural Networks" with Jana Leible, Mitja Nikolaus, David Rau, Dennis Ulmer, Tim Baumgärtner, and Elia Bruni accepted at 's workshop! Check out the preprint at

  14. 1. kol 2019.

    Dennis Ulmer explaining compositionality using "falling cows sign" example love it !!

  15. 3. ruj 2018.

    New paper at workshop with and Sanne Bouwmeester on analyzing how seq2seq models process disfluencies, using synthetic task-oriented dialogue data

  16. 24. srp 2019.
  17. 15. ruj 2018.

    A blog and preprint for our EMNLP paper "Interpreting Neural Networks With Nearest Neighbors" with . We address model overconfidence issues, leading to improvements in interpretability (1/2)

    Prikaži ovu nit
  18. 1. stu 2018.

    Leila Wehbe giving the third invited talk of

    Prikaži ovu nit
  19. 12. lip 2019.

    2019 paper is out! Neural LMs trained on English can suppress and recover syntactic expectations, approximating stack-like data structures; but recovery is imperfect! With

  20. 5. ruj 2018.

    Looking at some of the papers, I'm amazed but also a little concerned: imho many deserve a main conf slot. Did they get rejected from main confs and if so wtf is wrong with our field? 🙃

    Prikaži ovu nit

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.