Search papers, labs, and topics across Lattice.
This paper details the AI Wizards' approach to multimodal sexism identification in memes for the EXIST 2026 competition, utilizing a hierarchical model that employs conditional soft-label prediction based on empirical annotator distributions. The system leverages fixed Gemini Embedding 2 representations and a lightweight Gated MLP, optimized through KL divergence and homoscedastic uncertainty weighting. Notably, the submission achieved first place in Task 2.3 and ranked fourth in Tasks 2.1 and 2.2 on the official leaderboards, demonstrating the effectiveness of their method in a challenging domain.
Achieving top rankings in a competitive setting, this work reveals how hierarchical soft-label learning can significantly enhance multimodal sexism detection in memes.
We present the AI Wizards submission to EXIST 2026 for multimodal sexism identification in memes. The task is composed of three, increasingly harder subtasks. We model them hierarchically as conditional soft-label prediction over empirical annotator distributions. Our system maps fixed Gemini Embedding 2 vision-language representations through a lightweight Gated MLP trained with KL divergence and homoscedastic uncertainty weighting. Our submissions ranked first on Task 2.3 and fourth on Tasks 2.1 and 2.2 on the official Soft-Soft leaderboards. The code is available at https://github.com/NLP-AI-Wizards/EXIST-2026
ETH