[1909.02060] Distributionally Robust Language Modeling

Abstract:Language models are generally trained on data spanning a wide range of topics (e.g., news, reviews, fiction), but they might be applied to an a priori unknown target distribution (e.g., restaurant reviews). In this paper, we first show that training on text outside the test distribution can degrade test performance when using standard maximum likelihood (MLE) training. To remedy this without the knowledge of the test distribution, we propose an approach which trains a model that performs well over a wide range of potential test distributions. In particular, we derive a new distributionally robust optimization (DRO) procedure which minimizes the loss of the model over the worst-case mixture of topics with sufficient overlap with the training distribution. Our approach, called topic conditional value at risk (topic CVaR), obtains a 5.5 point perplexity reduction over MLE when the language models are trained on a mixture of Yelp reviews and news and tested only on reviews.

Distributionally Robust Language Modeling Yonatan Oren*1 Shiori Sagawa*1 Tatsunori B. Hashimoto*1,2 Percy Liang1 (* equal contribution) 1 Stanford Computer Science 2 Stanford Statistics {yonatano,thashim}@stanford.edu…

saved by

• Akira Yoshiyama

related reading

• [1911.08731] Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationarxiv.org
• Esoteric Language Modelsarxiv.org
• 2409.02908arxiv.org
• [2411.00640] Adding Error Bars to Evals: A Statistical Approach to Language Model Evaluationsarxiv.org
• [2305.18290] Direct Preference Optimization: Your Language Model is Secretly a Reward Modelarxiv.org
• [2605.12715] Scaling Laws for Mixture Pretraining Under Data Constraintsarxiv.org
• [2106.09685] LoRA: Low-Rank Adaptation of Large Language Modelsarxiv.org
• [2605.15220] Always Learning, Always Mixing: Efficient and Simple Data Mixing All The Timearxiv.org
• [2309.10668] Language Modeling Is Compressionarxiv.org
• [2606.04071] Covert Influence Between Language Modelsarxiv.org
• 2406.07524arxiv.org
• [2511.08579] Training Language Models to Explain Their Own Computationsarxiv.org