# BioSentinel at EXIST 2026: Soft-Label Optimization with XLM-RoBERTa for Sexism Intent Classification in Memes

arXiv: https://arxiv.org/abs/2607.24137 · HTML: https://mlsanity.com/paper/2607.24137/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 10 of its 20 references that are in the map.
- First posted on arXiv 27 Jul 2026.

## Builds on (most cited first, 10 of 10)

- [On Calibration of Modern Neural Networks](https://mlsanity.com/paper/1706.04599/) (2017)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [Learning from Noisy Labels with Deep Neural Networks: A Survey](https://mlsanity.com/paper/2007.08199/) (2020)
- [The Hateful Memes Challenge: Detecting Hate Speech in Multimodal Memes](https://mlsanity.com/paper/2005.04790/) (2020)
- [Dealing with Disagreements: Looking Beyond the Majority Vote in Subjective Annotations](https://mlsanity.com/paper/2110.05719/) (2021)
- [Human uncertainty makes classification more robust](https://mlsanity.com/paper/1908.07086/) (2019)
- [Label Distribution Learning](https://mlsanity.com/paper/1408.6027/) (2014)
- [Hate Speech Detection and Racial Bias Mitigation in Social Media based on BERT model](https://mlsanity.com/paper/2008.06460/) (2020)
- [SemEval-2023 Task 10: Explainable Detection of Online Sexism](https://mlsanity.com/paper/2303.04222/) (2023)
- [Re-examining Sexism and Misogyny Classification with Annotator Attitudes](https://mlsanity.com/paper/2410.03543/) (2024)

## Abstract (arXiv)

This paper describes the BioSentinel team's participation in EXIST 2026 Task 2.2: Source Intention in Memes, part of the CLEF 2026 evaluation campaign. The task requires classifying the communicative intent behind memes as direct, judgemental, or no (non-sexist), under a Learning with Disagreement (Le-Wi-Di) paradigm that mandates both hard-label and soft-label (probability distribution) predictions. We present a text-centric approach built on xlm-roberta-base (270M parameters) trained with a composite loss function combining KL divergence on soft annotator distributions and weighted cross-entropy on hard labels. On the official test set, the system achieved an ICM-Soft-Norm of 0.3229 and ICM-Norm of 0.3778, with a hard F1-score of 0.4236, ranking 40th (out of 118 submissions) in the soft-soft evaluation and 49th (out of 187 submissions) in the hard-hard evaluation. We provide an analysis of the dataset characteristics, exploratory larger-architecture runs, and the role of annotator disagreement in shaping model design for subjective NLP tasks. Ablation results show that KL loss improves soft-label metrics, while CE loss improves hard-label accuracy. We also report a separate validation-set temperature analysis.