NLP Wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Distribution Shift ======
Also known as **covariate shift** (shift between training and testing in the distribution of the input variables), see [[https://jmlr.csail.mit.edu/papers/volume10/bickel09a/bickel09a.pdf|Bickel 2009]].

===== Papers =====
  * [[https://jmlr.csail.mit.edu/papers/volume10/bickel09a/bickel09a.pdf|Bickel et al 2009 - Discriminative Learning Under Covariate Shift]]
  * [[https://arxiv.org/pdf/1810.08750.pdf|Duchi 2018 - Learning Models with Uniform Performance via Distributionally Robust Optimization]]
  * [[https://arxiv.org/pdf/2007.13982.pdf|Duchi et al 2020 - Distributionally Robust Losses for Latent Covariate Mixtures]]
  * [[https://proceedings.mlr.press/v162/zhou22d/zhou22d.pdf|Zhou et al 2022 - Model Agnostic Sample Reweighting for Out-of-Distribution Learning]]

==== NLP ====
  * [[https://arxiv.org/pdf/2103.10282.pdf|Michel et al 2021 - Modeling the Second Player in Distributionally Robust Optimization]]
  * [[https://arxiv.org/pdf/2109.01558.pdf|Michel 2021 - Learning Neural Models for Natural Language Processing in the Face of Distributional Shift]] PhD thesis
  * [[https://arxiv.org/pdf/2204.06340.pdf|Michel 2022 - Distributionally Robust Models with Parametric Likelihood Ratios]]

===== Distribution Shift / Out-of-Domain Detection =====
  * **In NLP**
    * [[https://aclanthology.org/2023.acl-long.717.pdf|Uppaal et al 2023 - Is Fine-tuning Needed? Pre-trained Language Models Are Near Perfect for Out-of-Domain Detection]]

===== Datasets =====
  * WILDS: ([[https://arxiv.org/pdf/2012.07421.pdf|paper]]) A benchmark of in-the-wild distribution shifts

===== People =====
  * [[https://scholar.google.com/citations?user=oyyIf0YAAAAJ&hl=en|Paul Michel]] See his [[https://arxiv.org/pdf/2109.01558.pdf|2021 thesis]]
  * [[https://scholar.google.com/citations?user=Nn990CkAAAAJ&hl=en|Pang Wei Koh]]
  * [[https://scholar.google.com/citations?user=pouyVyUAAAAJ&hl=en|Percy Liang]]
  * [[https://scholar.google.com/citations?user=5ygiTwsAAAAJ&hl=en|Tatsunori Hashimoto]]

===== Related Pages =====
  * [[Fairness]] (Methods that are not robust to distribution shift may not be fair across populations)
  * [[nlp:Robustness in NLP]]