JusticeBots@LT-EDI 2026: Prompt-Based Counter-Narrative Generation for Homophobia and Transphobia Comments

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Public Safety & Security · Depth: Expert, quick

Summary

The JusticeBots@LT-EDI 2026 team participated in a shared task focused on generating counter-narratives for homophobic and transphobic online comments. Addressing the rise of hate speech targeting LGBTQ+ individuals, their approach utilized zero-shot prompting with large language models, specifically GPT-4o, Gemini 1.5 Pro, and Llama-3 Sonar Large via Perplexity AI. Instead of training a dedicated model, the team designed a structured prompt to guide these publicly available AI tools in producing respectful, concise, and contextually appropriate responses. Experiments conducted on English and Tamil comments demonstrated that this prompt-based method successfully generated meaningful multilingual counter-narratives without requiring additional training, highlighting LLMs' potential as lightweight tools for counter-speech in diverse online environments.

Key takeaway

For NLP Engineers developing anti-hate speech tools, you should prioritize prompt engineering with readily available large language models like GPT-4o or Gemini 1.5 Pro. This approach allows for rapid deployment of multilingual counter-narrative generation without extensive model training, significantly reducing development overhead and time-to-market for online content moderation solutions. Focus on crafting precise prompts to achieve desired output characteristics.

Key insights

Prompt-based zero-shot LLM generation effectively creates multilingual counter-narratives against online hate speech.

Principles

• Counter-narratives promote empathy and factual clarity.
• Zero-shot prompting can replace task-specific model training.
• Structured prompts guide LLMs for specific outputs.

Method

Design a structured prompt to guide large language models (GPT-4o, Gemini 1.5 Pro, Llama-3 Sonar Large) for generating respectful, concise, and contextually appropriate counter-narratives in a zero-shot manner.

In practice

• Use GPT-4o, Gemini 1.5 Pro for counter-speech.
• Apply zero-shot prompting for multilingual content.
• Craft structured prompts for desired tone/style.

Topics

• Counter-Narrative Generation
• Zero-Shot Prompting
• Large Language Models
• Online Hate Speech
• Multilingual AI
• Prompt Engineering

Best for: AI Engineer, Machine Learning Engineer, Research Scientist, AI Scientist, NLP Engineer, Prompt Engineer

Related on AIssential

• Automatic Generation of Highlights for Academic Paper Via Prompt-based Learning — Prompt-based learning with LLMs can generate academic paper highlights effectively without extensive labeled data.
• Announcing the ICLR 2026 Outstanding Papers — Transformer efficiency stems from succinct concept encoding, while LLMs struggle with multi-turn conversations due to training data mismatches.
• Grab the 5 prompts that get you ready for Fable 5 before it's back — Fable 5, a 10-trojan parameter model, sets a benchmark for future large language models, necessitating new interaction strategies.
• Guarantee zero “hallucinations” for your RAG Agent!* — Guaranteeing RAG "hallucination-free" output requires strict context grounding, trading abstractive freedom for verifiable traceability.
• LLM-based Detection of Manipulative Political Narratives — LLM-driven pipelines can effectively identify and cluster manipulative political narratives in raw social media data.
• Locale-Conditioned Few-Shot Prompting Mitigates Demonstration Regurgitation in On-Device PII Substitution with Small Language Models — Locale-conditioned few-shot prompting prevents SLMs from regurgitating PII substitution demonstrations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.