Igniters@LTEDI 2026: Multilingual Gender-Inclusive Language Generation with mT5 and Counter-Narrative Generation Using Llama-3

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

Summary

The Igniters@LTEDI 2026 framework introduces a dual-paradigm approach to mitigate societal stereotypes propagated by Large Language Models. It specifically targets multilingual gender-inclusivity and counter-narrative generation. For gender-neutral text transformation, a fine-tuned mT5 encoder-decoder model performs controlled sentence rewriting, ensuring minimal edits while preserving semantic fidelity and grammatical fluency. The counter-narrative component employs the Llama-3 8B decoder-only model, generating empathetic and persuasive responses through structured prompt-based generation. Evaluated using LT-EDI ACL 2026 shared task datasets across English, Tamil, Kannada, German, and Spanish, the framework demonstrates strong effectiveness. It successfully identifies and neutralizes gender markers, particularly in morphologically rich languages, and achieves high performance in politeness, coherence, and relevance for counter-narratives, contributing to responsible multilingual NLP systems.

Key takeaway

For NLP Engineers developing multilingual systems, this framework offers a robust method to address gender bias and generate empathetic counter-narratives. You should consider fine-tuning mT5 for controlled gender-neutral text transformation, especially for morphologically rich languages. Additionally, employing Llama-3 8B with structured prompts can effectively produce polite, coherent, and relevant counter-narratives, enhancing the responsibility and inclusivity of your deployed LLM applications.

Key insights

A dual-paradigm framework uses mT5 and Llama-3 to achieve multilingual gender-inclusive and counter-narrative language generation.

Principles

• LLMs propagate societal stereotypes.
• Controlled rewriting preserves semantic fidelity.
• Structured prompting yields empathetic responses.

Method

Fine-tune mT5 for controlled sentence rewriting to neutralize gender markers. Employ Llama-3 8B with structured prompts for empathetic, persuasive counter-narrative generation.

In practice

• Fine-tune mT5 for gender-neutral text.
• Use Llama-3 8B for empathetic responses.
• Apply structured prompts for LLM control.

Topics

• Multilingual NLP
• Gender Bias Mitigation
• Counter-Narrative Generation
• mT5
• Llama-3
• Responsible AI

Best for: Research Scientist, AI Scientist, NLP Engineer, AI Ethicist

Related on AIssential

• Bias and Justice in Language Models: Evidence from a Linguistically, Socially, and Culturally Asymmetric Literature — LLM bias research is skewed towards English and binary gender, lacking intersectional and multilingual real-world analyses.
• Mod-Guide: An LLM-based Content Moderation Feedback System to Address Insensitive Speech toward Indigenous Ethnic and Religious Minority Communities — Integrating culturally grounded narratives via RAG enhances LLM sensitivity to minority perspectives in content moderation.
• Extending the Contact Hypothesis: Cross-Linguistic Evaluation of Religion and Nationality Bias When Prompting LLMs in German and Icelandic — LLMs exhibit religion and nationality biases, requiring cross-linguistic evaluation for fairness.
• MORPHOGEN: A Multilingual Benchmark for Evaluating Gender-Aware Morphological Generation — MORPHOGEN evaluates LLM gender-aware morphological generation in French, Arabic, and Hindi via a sentence rewriting task.
• Location Not Found: Exposing Implicit Local and Global Biases in Multilingual LLMs — Multilingual LLMs exhibit implicit US-centric and population-driven biases, exacerbated by instruction tuning.
• IndoBias: A Dual Track Culturally Grounded Benchmark for LLMs Bias Evaluation in Indonesian Languages — Culturally-grounded benchmarks are crucial for evaluating LLM bias in diverse, multilingual contexts like Indonesia.

Open in AIssential →

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