hate speech

Permalink ArXiv Audio Speech

"LLMs-Integrated Automatic Hate Speech Recognition Using Controllable Text Generation Models"

ArXiv Audio Speech

* Cited for critical analysis under Article 32.

MultiHateLoc: AI for Temporal Localization of Hate Speech in Videos

Research #Hate Speech 🔬 Research|Analyzed: Jan 10, 2026 12:04•

Published: Dec 11, 2025 08:18

•

1 min read

•ArXiv

Analysis

This research paper explores the challenging problem of identifying and locating hate speech within online videos using multimodal AI. The work likely contributes to advancements in content moderation and online safety by offering a technical solution for detecting harmful content.

Key Takeaways

•Focuses on identifying hate speech in video format.
•Utilizes multimodal data (e.g., audio, video, text).
•Aims to improve online content moderation.

Reference / Citation

"The paper focuses on the temporal localization of multimodal hate content."

* Cited for critical analysis under Article 32.

Reassessing LLM Reliability: Can Large Language Models Accurately Detect Hate Speech?

Research #LLM 🔬 Research|Analyzed: Jan 10, 2026 12:19•

Published: Dec 10, 2025 14:00

•

1 min read

•ArXiv

Analysis

This research explores the limitations of Large Language Models (LLMs) in detecting hate speech, focusing on their ability to evaluate concepts they might not be able to fully annotate. The study likely examines the implications of this disconnect on the reliability of LLMs in crucial applications.

Key Takeaways

•LLMs might struggle to accurately detect hate speech when relying on evaluations of concepts they can't annotate.
•The research likely investigates how this limitation affects the overall reliability of LLMs.
•The findings will have implications for the deployment of LLMs in applications requiring accurate content moderation.

Reference / Citation

"The study investigates LLM reliability in the context of hate speech detection."

* Cited for critical analysis under Article 32.

Fine-Grained Chinese Hate Speech Detection: A Prompt-Driven LLM Merge Approach

Research #LLM 🔬 Research|Analyzed: Jan 10, 2026 12:21•

Published: Dec 10, 2025 11:58

•

1 min read

•ArXiv

Analysis

This research explores merging large language models (LLMs) to enhance fine-grained hate speech detection in Chinese, a crucial area for mitigating online toxicity. The work's reliance on prompt engineering for the merged LLMs warrants further investigation into its robustness and generalizability across diverse data distributions.

Key Takeaways

•Investigates the use of LLM merging for hate speech detection.
•Focuses specifically on the Chinese language and fine-grained detection.
•Employs a prompt-driven approach for the merged models.

Reference / Citation

"The study focuses on prompt-driven LLM merge for fine-grained Chinese hate speech detection."

* Cited for critical analysis under Article 32.

Reasoning-Aware Multimodal Fusion for Hateful Video Detection

Research #Video Detection 🔬 Research|Analyzed: Jan 10, 2026 13:28•

Published: Dec 2, 2025 13:24

•

1 min read

•ArXiv

Analysis

This article, sourced from ArXiv, likely details a research paper proposing a novel approach to detecting hateful content in videos. The focus on 'Reasoning-Aware Multimodal Fusion' suggests an innovative combination of different data modalities and reasoning capabilities for improved accuracy.

Key Takeaways

•The research likely explores combining visual, audio, and textual information within videos to detect hate speech.
•The 'reasoning-aware' aspect probably involves employing AI to understand the context and intent behind the video content.
•The study could provide insights into creating more robust and accurate hate speech detection systems.

Reference / Citation

"The article's context indicates the subject matter focuses on hateful video detection using multimodal data fusion and reasoning."

* Cited for critical analysis under Article 32.

Feature Selection Boosts BERT for Hate Speech Detection

Research #Hate Speech 🔬 Research|Analyzed: Jan 10, 2026 13:35•

Published: Dec 1, 2025 19:11

•

1 min read

•ArXiv

Analysis

This research explores enhancements to BERT for hate speech detection, a critical area in AI safety and online content moderation. The vocabulary augmentation aspect suggests an attempt to improve robustness against variations in language and slang.

Key Takeaways

•Applies feature selection techniques to improve BERT's performance in hate speech detection.
•Employs vocabulary augmentation to enhance the model's ability to recognize varied language.
•Contributes to the ongoing effort of making AI systems safer and more reliable in content analysis.

Reference / Citation

"The study focuses on using Feature Selection and Vocabulary Augmentation with BERT to detect hate speech."

* Cited for critical analysis under Article 32.

Gradient Masters Tackle Bengali Hate Speech: Advancing Low-Resource NLP

Research #NLP 🔬 Research|Analyzed: Jan 10, 2026 14:26•

Published: Nov 23, 2025 07:29

•

1 min read

•ArXiv

Analysis

This research paper focuses on a critical challenge: detecting hate speech in a low-resource language. The use of ensemble-based adversarial training is a promising approach to improve model robustness and accuracy in this context.

Key Takeaways

•The paper explores hate speech detection in Bengali, a low-resource language.
•The methodology employs ensemble-based adversarial training.
•The research contributes to improved NLP for under-resourced languages.

Reference / Citation

"The research focuses on the BLP-2025 Task 1, addressing hate speech detection."

* Cited for critical analysis under Article 32.