Beyond Binary Classification: Detecting Fine-Grained Sexism in Social Media Videos
arXiv:2602.15757v1 Announce Type: new Abstract: Online sexism appears in various forms, which makes its detection challenging. Although automated tools can enhance the identification of sexist content, they are often restricted to binary classification. Consequently, more subtle manifestations of sexism may remain undetected due to the lack of fine-grained, context-sensitive labels. To address this issue, we make the following contributions: (1) we present FineMuSe, a new multimodal sexism detection dataset in Spanish that includes both binary and fine-grained annotations; (2) we introduce a comprehensive hierarchical taxonomy that encompasses forms of sexism, non-sexism, and rhetorical devices of irony and humor; and (3) we evaluate a wide range of LLMs for both binary and fine-grained sexism detection. Our findings indicate that multimodal LLMs perform competitively with human annotators in identifying nuanced forms of sexism; however, they struggle to capture co-occurring sexist ty
arXiv:2602.15757v1 Announce Type: new Abstract: Online sexism appears in various forms, which makes its detection challenging. Although automated tools can enhance the identification of sexist content, they are often restricted to binary classification. Consequently, more subtle manifestations of sexism may remain undetected due to the lack of fine-grained, context-sensitive labels. To address this issue, we make the following contributions: (1) we present FineMuSe, a new multimodal sexism detection dataset in Spanish that includes both binary and fine-grained annotations; (2) we introduce a comprehensive hierarchical taxonomy that encompasses forms of sexism, non-sexism, and rhetorical devices of irony and humor; and (3) we evaluate a wide range of LLMs for both binary and fine-grained sexism detection. Our findings indicate that multimodal LLMs perform competitively with human annotators in identifying nuanced forms of sexism; however, they struggle to capture co-occurring sexist types when these are conveyed through visual cues.
Executive Summary
This article presents FineMuSe, a new multimodal sexism detection dataset in Spanish that includes both binary and fine-grained annotations. The authors evaluate a wide range of Large Language Models (LLMs) for both binary and fine-grained sexism detection, demonstrating that multimodal LLMs perform competitively with human annotators in identifying nuanced forms of sexism. However, they struggle to capture co-occurring sexist types when these are conveyed through visual cues. The study contributes to the development of more sophisticated automated tools for detecting sexism in social media videos and highlights the importance of context-sensitive labels for nuanced sexism detection. The findings have significant implications for the design and implementation of AI-powered sexism detection systems and the development of more effective strategies for mitigating online sexism.
Key Points
- ▸ Introduction of FineMuSe, a multimodal sexism detection dataset in Spanish with binary and fine-grained annotations
- ▸ Evaluation of LLMs for binary and fine-grained sexism detection
- ▸ Demonstration of the effectiveness of multimodal LLMs in identifying nuanced forms of sexism
Merits
Strength in methodology
The authors employ a comprehensive hierarchical taxonomy that encompasses forms of sexism, non-sexism, and rhetorical devices of irony and humor, providing a robust framework for evaluating LLMs and detecting nuanced forms of sexism.
Significance of findings
The study highlights the importance of context-sensitive labels for nuanced sexism detection and demonstrates the potential of multimodal LLMs in identifying subtle manifestations of sexism.
Demerits
Limitation in visual cue detection
The authors note that multimodal LLMs struggle to capture co-occurring sexist types when these are conveyed through visual cues, highlighting the need for further research in this area.
Dataset limitations
The study is limited by the availability of a single dataset (FineMuSe) in Spanish, which may not generalize to other languages or cultural contexts.
Expert Commentary
The study represents a significant contribution to the field of sexism detection in social media videos. The authors' use of a comprehensive hierarchical taxonomy and evaluation of LLMs demonstrates a deep understanding of the complexities of sexism and the limitations of existing automated tools. However, the study's findings also highlight the need for further research in the area of visual cue detection, particularly in diverse cultural and linguistic contexts. The study's implications for AI-powered sexism detection systems and policy strategies for mitigating online sexism are substantial, and it is essential to build on this research to develop more effective solutions for addressing online sexism.
Recommendations
- ✓ Future research should focus on developing multimodal LLMs that can capture co-occurring sexist types conveyed through visual cues.
- ✓ The development of more robust evaluation and benchmarking frameworks for LLMs is essential for ensuring their safety and reliability in real-world applications.
