Self-Correction for OOD Generalization
Abstract
In this work, we aim to study how the self-correction mechanisms aid OOD (out-of-distribution) generalization in both multimodal and language-only models. Reasoning based methods like self-refine and STaR have helped to improve the correction capacity of the language models; however there have been no studies quantifying the reasoning improvement to help OOD generalization of these models. Initial results, show an improvement of 1.6%-2% on an OOD dataset where the model is finetuned using either self-refinement or STaR on an ID (in-distribution) dataset.
Cite
Text
Kumar et al. "Self-Correction for OOD Generalization." ICLR 2025 Workshops: SSI-FM, 2025.Markdown
[Kumar et al. "Self-Correction for OOD Generalization." ICLR 2025 Workshops: SSI-FM, 2025.](https://mlanthology.org/iclrw/2025/kumar2025iclrw-selfcorrection/)BibTeX
@inproceedings{kumar2025iclrw-selfcorrection,
title = {{Self-Correction for OOD Generalization}},
author = {Kumar, Vanya Bannihatti and Rao, Abhinav Sukumar and Raghunathan, Aditi},
booktitle = {ICLR 2025 Workshops: SSI-FM},
year = {2025},
url = {https://mlanthology.org/iclrw/2025/kumar2025iclrw-selfcorrection/}
}