Jiaqian Ren

Dual-Path Counterfactual Integration for Multimodal Aspect-Based Sentiment Classification

Multimodal aspect-based sentiment classification (MABSC) requires fine-grained reasoning over both textual and visual content to infer sentiments toward specific aspects. However, existing methods often rely on superficial correlations—particularly between aspect terms and sentiment labels—leading t