paper-with-me

Papers

Generalizing to any diverse distribution: uniformity, gentle finetuning and rebalancing

2024-10-08 · Andreas Loukas, Karolis Martinkus, Ed Wagstaff, Kyunghyun Cho

As training datasets grow larger, we aspire to develop models that generalize well to any diverse test distribution, even if the latter deviates significantly from the training data. Various approaches like domain adaptation, domain generalization, and robust optimization attempt to address the out-of-distribution challenge by posing assumptions about the relation between training and test distribution. Differently, we adopt a more conservative perspective by accounting for the worst-case error across all sufficiently diverse test distributions within a known domain. Our first finding is that training on a uniform distribution over this domain is optimal. We also interrogate practical remedies when uniform samples are unavailable by considering methods for mitigating non-uniformity through finetuning and rebalancing. Our theory provides a mathematical grounding for previous observations on the role of entropy and rebalancing for o.o.d. generalization and foundation model training. We also provide new empirical evidence across tasks involving o.o.d. shifts which illustrate the broad applicability of our perspective.

📄 PDF Abstract BibTeX arXiv:2410.05980

Code (0)

등록된 구현이 없습니다.

Tasks

Domain AdaptationDomain Generalization

Similar Papers 제목 키워드 기반

GENTLE: A Genre-Diverse Multilayer Challenge Set for English NLP and Linguistic Evaluation

2023-06-03 · Tatsuya Aoyama, Shabnam Behzad, Luke Gessler, Lauren Levine 외

We present GENTLE, a new mixed-genre English challenge corpus totaling 17K tokens and consisting of 8 unusual text types for out-of domain evaluation: dictionary entries, esports commentaries, legal documents, medical no…

coreference-resolutionCoreference ResolutionDependency ParsingDiscourse Parsing+1

Learning Dynamics of VLM Finetuning

2025-10-13 · Jusheng Zhang, Kaitong Cai, Jing Yang, Keze Wang arxiv

Preference-based finetuning of vision--language models (VLMs) is brittle: trivially wrong negatives inject uninformative gradients that destabilize training. We recast alignment as \textbf{learning-dynamics--aware optimi…

RAU: Towards Regularized Alignment and Uniformity for Representation Learning in Recommendation

2025-03-24 · Xi Wu, Dan Zhang, Chao Zhou, Liangwei Yang 외

Recommender systems (RecSys) have become essential in modern society, driving user engagement and satisfaction across diverse online platforms. Most RecSys focuses on designing a powerful encoder to embed users and items…

Recommendation SystemsRepresentation Learning

Generalizing and Decoupling Neural Collapse via Hyperspherical Uniformity Gap

2023-03-11 · Weiyang Liu, Longhui Yu, Adrian Weller, Bernhard Schölkopf

The neural collapse (NC) phenomenon describes an underlying geometric symmetry for deep neural networks, where both deeply learned features and classifiers converge to a simplex equiangular tight frame. It has been shown…

Mr Darcy and Mr Toad, gentlemen: distributional names and their kinds

2015-04-01 · WS 2015 4 · Aur{\'e}lie Herbelot