We are delighted to announce that a paper from the Intelligent Information Processing Lab (IIPL) has been accepted to the 2025 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2025).

Title: 

See-Saw Modality Balance: See Gradient, and Sew Impaired Vision-Language Balance to Mitigate Dominant Modality Bias.

Authors:

Junehyoung Kwon, MiHyeon Kim, Eunju Lee, Juhwan Choi, YoungBin Kim

Abstract:

Vision-language (VL) models have demonstrated strong performance across various tasks. However, these models often rely on a specific modality for predictions, leading to "dominant modality bias." This bias significantly degrades performance, especially when one modality is impaired. In this study, we analyze model behavior under dominant modality bias and theoretically show that unaligned gradients or differences in gradient magnitudes prevent balanced convergence of the loss. Based on these findings, we propose a novel framework, BalGrad, to mitigate dominant modality bias. Our approach includes inter-modality gradient reweighting, adjusting the gradient of KL divergence based on each modality's contribution, and inter-task gradient projection to align task directions in a non-conflicting manner. Experiments on UPMC Food-101, Hateful Memes, and MM-IMDb datasets confirm that BalGrad effectively alleviates over-reliance on specific modalities, improving model robustness across diverse vision-language tasks.