Evaluating Generative AI Beyond Comprehension: A G11n Perspective

Track: Technical | T1 |   Everyone |
Wednesday, October 15, 2025, 9:00am – 9:30am
Held in: Steinbeck 3
Presenters:
Edgar Castillo - Dilato Infotech Limited 
Patricia Oceguera - Autonomous University of Baja California
Host: Daniel Goldschmidt

We have identified gloablization (G11n) bugs in popular LLMs, revealing challenges in AI’s adaptation to diverse languages and cultures. As AI grows globally, assessing content quality becomes increasingly complex. To tackle this, we created the GenAI G11n Assessment Model, a framework to evaluate AI performance beyond comprehension. It ensures accuracy, inclusivity, and cultural relevance in multilingual interactions. This structured approach addresses the variability of AI-generated content and improves adaptability across contexts. Join us to discuss how this model can enhance AI evaluation and make global AI systems more effective and inclusive.

Key Takeaways:

A clear understanding of the challenges AI faces when adapting to different languages and cultures; a practical framework (the GenAI G11n Assessment Model) to evaluate AI performance for global use; we will release the proprietary dataset used in the model for all the attendees.