Multimodal Negative Learning

TL;DR

This paper introduces a novel multimodal learning paradigm called 'Learning Not to be' that uses negative guidance to improve robustness and preserve modality-specific information, especially in noisy or imbalanced settings.

Contribution

It proposes the Multimodal Negative Learning framework, a new approach that enhances robustness by dynamically guiding weak modalities to suppress non-target classes, avoiding over-alignment.

Findings

Tightens the robustness lower bound of multimodal learning.

Reduces empirical error of weak modalities under noise and imbalance.

Demonstrates effectiveness across multiple benchmarks.

Abstract

Multimodal learning systems often encounter challenges related to modality imbalance, where a dominant modality may overshadow others, thereby hindering the learning of weak modalities. Conventional approaches often force weak modalities to align with dominant ones in "Learning to be (the same)" (Positive Learning), which risks suppressing the unique information inherent in the weak modalities. To address this challenge, we offer a new learning paradigm: "Learning Not to be" (Negative Learning). Instead of enhancing weak modalities' target-class predictions, the dominant modalities dynamically guide the weak modality to suppress non-target classes. This stabilizes the decision space and preserves modality-specific information, allowing weak modalities to preserve unique information without being over-aligned. We proceed to reveal multimodal learning from a robustness perspective and theoretically derive the Multimodal Negative Learning (MNL) framework, which introduces a dynamic guidance mechanism tailored for negative learning. Our method provably tightens the robustness lower bound of multimodal learning by increasing the Unimodal Confidence Margin (UCoM) and reduces the empirical error of weak modalities, particularly under noisy and imbalanced scenarios. Extensive experiments across multiple benchmarks demonstrate the effectiveness and generalizability of our approach against competing methods. The code will be available at https://github.com/BaoquanGong/Multimodal-Negative-Learning.git.