Video by Open Data Science and AI Conference via YouTube
LLMs can solve Olympiad-level math, complex coding problems, and expert benchmarks—but still fail in surprisingly simple ways.
In this ODSC AI East 2026 keynote, Nouha Dziri, PhD, Senior Research Scientist at Cohere Labs, explores what it takes to build more trustworthy large language models. Drawing on her research in reasoning, post-training, and AI safety, Nouha examines the limits of current LLMs, why “jagged intelligence” creates real-world reliability challenges, and how better training and evaluation can improve model behavior.
The session covers key challenges in LLM deployment, from reasoning failures and out-of-distribution generalization to reinforcement learning approaches that strengthen capabilities and scalable safety systems like automated adversarial testing and moderation.
Watch to learn how researchers are working to advance LLM capabilities while making them safer, more robust, and more reliable in practice.
————————————————————————————————————-
Visit our website and choose the nearest ODSC event to attend and experience all our training and workshops: https://odsc.ai
To watch more videos like this, visit https://aiplus.training
Sign up for the newsletter to stay up to date with the latest trends in data science: https://opendatascience.com/newsletter/
Follow us online!
• Facebook: https://www.facebook.com/OPENDATASCI
• Instagram: https://www.instagram.com/odsc/
• Blog: https://opendatascience.com/
• LinkedIn: https://www.linkedin.com/company/open-data-science/
• X (twitter): https://x.com/_odsc