See, Explain, and Intervene: A Few-Shot Multimodal Agent Framework for Hateful Meme Moderation
Naquee Rizwan, Subhankar Swain, Paramananda Bhaskar, Gagan Aryan, Shehryaar Shah Khan, Animesh Mukherjee {nrizwan, subhankarswain.24, pbhaskar, khanshehryaar705}@[email protected], [email protected] · 2026
Read See, Explain, and Intervene: A Few-Shot Multimodal Agent Framework for Hateful Meme Moderation by Naquee Rizwan, Subhankar Swain, Paramananda Bhaskar, Gagan Aryan, Shehryaar Shah Khan, Animesh Mukherjee {nrizwan, subhankarswain.24, pbhaskar, khanshehryaar705}@[email protected], [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-04692.txt.
Description
Naquee Rizwan and five co-authors tackle hateful memes as a moderation pipeline rather than a single classifier, in a 2026 arXiv paper. Real moderation needs three things — detecting the hateful meme, explaining why it is hateful, and intervening before posting. Yet research usually studies detection alone, and labelled meme data is too expensive to gather at scale. Their framework, PEST, is built for that scarcity.
PEST steers black-box vision-language models without opening them: task-specific generative models and few-shot prompting adapt the large models to different meme types, with detection, explanation, and intervention handled as one workflow. The authors describe the system as, to their knowledge, the first aimed at generalizable hateful meme moderation under limited data, and argue it is shaped for real production deployment. The abstract carries the authors' own warning that the paper contains potentially toxic example content, quoted for research purposes.
Catalog record
| CREATOR | Naquee Rizwan, Subhankar Swain, Paramananda Bhaskar, Gagan Aryan, Shehryaar Shah Khan, Animesh Mukherjee {nrizwan, subhankarswain.24, pbhaskar, khanshehryaar705}@[email protected], [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 40 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-04692