MEME-Fusion@CHiPSAL 2026: Multimodal Ablation Study of Hate Detection and Sentiment Analysis on Nepali Memes

TL;DR

This paper introduces a multimodal fusion architecture for hate speech and sentiment detection in Nepali memes, demonstrating improved performance and revealing key challenges in low-resource, script-specific contexts.

Contribution

It proposes a hybrid cross-modal attention model combining visual and multilingual text encoders, with insights into model limitations and data scarcity effects.

Findings

Explicit cross-modal reasoning improves F1-macro by 5.9% over text-only baselines.

English-centric vision models perform poorly on Devanagari script.

Ensemble methods can degrade under data scarcity due to overfitting.

Abstract

Hate speech detection in Devanagari-scripted social media memes presents compounded challenges: multimodal content structure, script-specific linguistic complexity, and extreme data scarcity in low-resource settings. This paper presents our system for the CHiPSAL 2026 shared task, addressing both Subtask A (binary hate speech detection) and Subtask B (three-class sentiment classification: positive, neutral, negative). We propose a hybrid cross-modal attention fusion architecture that combines CLIP (ViT-B/32) for visual encoding with BGE-M3 for multilingual text representation, connected through 4-head self-attention and a learnable gating network that dynamically weights modality contributions on a per-sample basis. Systematic evaluation across eight model configurations demonstrates that explicit cross-modal reasoning achieves a 5.9% F1-macro improvement over text-only baselines on Subtask A, while uncovering two unexpected but critical findings: English-centric vision models exhibit near-random performance on Devanagari script, and standard ensemble methods catastrophically degrade under data scarcity (N nearly equal to 850 per fold) due to correlated overfitting. The code can be accessed at https://github.com/Tri-Yantra-Technologies/MEME-Fusion/