Summary of Towards Explainable Harmful Meme Detection Through Multimodal Debate Between Large Language Models, by Hongzhan Lin et al.
Towards Explainable Harmful Meme Detection through Multimodal Debate between Large Language Models
by Hongzhan Lin, Ziyang Luo, Wei Gao, Jing Ma, Bo Wang, Ruichao Yang
First submitted to arxiv on: 24 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This paper proposes an explainable approach to detect harmful memes, which is essential in today’s social media landscape. The proposed method uses Large Language Models (LLMs) to generate explanations derived from contradictory arguments between harmless and harmful positions. A small language model serves as a “debate judge” for fine-tuning the LLMs’ outputs. This allows the model to perform dialectical reasoning over intricate harm-indicative patterns, utilizing multimodal explanations. The approach is tested on three public meme datasets, demonstrating improved performance compared to state-of-the-art methods and providing better explainability of predictions. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us understand how to find harmful memes online. It’s a problem because these memes can be mean or hurtful. Right now, there are no good ways to detect them because they don’t always say what they mean. The researchers came up with a new way to do this by using special computer models that can talk and reason like humans. These models have debates about whether something is harmless or harmful. Then, another model helps figure out why it’s one or the other. This approach works really well on real-life meme datasets and gives good reasons for why it thinks something is harmful. |
Keywords
» Artificial intelligence » Fine tuning » Language model
