From Chain-of-Thought to Agentic AI: Demystifying Reasoning in LLMs

Resumen

Reasoning is one of the most powerful—and sometimes surprising—capabilities in large language models. In this talk, we’ll demystify what “reasoning” means, how intermediate steps improve problem-solving, and why certain techniques—like Chain-of-Thought, Self-Consistency, RL-based reasoning, and distilled reasoning models—can unlock smarter behavior. We’ll also explore how reasoning can emerge naturally through reinforcement learning. Through practical examples in generative and agentic AI, we’ll cover when reasoning delivers value, its trade-offs, and how to optimize context for better performance. Attendees will leave with a clear framework for applying reasoning models effectively.

Ponentes