# Gradient Masters at BLP-2025 Task 1: Advancing Low-Resource NLP for Bengali using Ensemble-Based Adversarial Training for Hate Speech Detection

arXiv: https://arxiv.org/abs/2511.18324 · HTML: https://mlsanity.com/paper/2511.18324/

## 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 23 Nov 2025.

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

- [Explaining and Harnessing Adversarial Examples](https://mlsanity.com/paper/1412.6572/) (2014)
- [Attacking Large Language Models with Projected Gradient Descent](https://mlsanity.com/paper/2402.09154/) (2024)
- [Hate Speech detection in the Bengali language: A dataset and its baseline evaluation](https://mlsanity.com/paper/2012.09686/) (2020)
- [Classification Benchmarks for Under-resourced Bengali Language based on Multichannel Convolutional-LSTM Network](https://mlsanity.com/paper/2004.07807/) (2020)
- [BD-SHS: A Benchmark Dataset for Learning to Detect Online Bangla Hate Speech in Different Social Contexts](https://mlsanity.com/paper/2206.00372/) (2022)
- [BanTH: A Multi-label Hate Speech Detection Dataset for Transliterated Bangla](https://mlsanity.com/paper/2410.13281/) (2024)
- [LLM-Based Multi-Task Bangla Hate Speech Detection: Type, Severity, and Target](https://mlsanity.com/paper/2510.01995/) (2025)
- [Improving robustness of language models from a geometry-aware perspective](https://mlsanity.com/paper/2204.13309/) (2022)
- [nlpBDpatriots at BLP-2023 Task 1: A Two-Step Classification for Violence Inciting Text Detection in Bangla](https://mlsanity.com/paper/2311.15029/) (2023)
- [BIDWESH: A Bangla Regional Based Hate Speech Detection Dataset](https://mlsanity.com/paper/2507.16183/) (2025)

## Abstract (arXiv)

This paper introduces the approach of "Gradient Masters" for BLP-2025 Task 1: "Bangla Multitask Hate Speech Identification Shared Task". We present an ensemble-based fine-tuning strategy for addressing subtasks 1A (hate-type classification) and 1B (target group classification) in YouTube comments. We propose a hybrid approach on a Bangla Language Model, which outperformed the baseline models and secured the 6th position in subtask 1A with a micro F1 score of 73.23% and the third position in subtask 1B with 73.28%. We conducted extensive experiments that evaluated the robustness of the model throughout the development and evaluation phases, including comparisons with other Language Model variants, to measure generalization in low-resource Bangla hate speech scenarios and data set coverage. In addition, we provide a detailed analysis of our findings, exploring misclassification patterns in the detection of hate speech.