Stanford CS25: V5 I Large Language Model Reasoning, Denny Zhou of Google Deepmind
- Source
- youtube.com
- Author
- Stanford Online
- Date
Context
In a Stanford CS25 lecture, Denny Zhou of Google DeepMind avoids debating whether models can reason and instead defines the term narrowly: the intermediate , meaning pieces of text, that a model writes between a question and its final answer. A task Zhou designed makes the point. Asked for the last letters of two words joined together, a model that answers directly tends to fail, while one that first writes out each word's last letter can succeed.
Zhou's main claim is that models trained only on text prediction can already produce such reasoning, and that the problem is decoding. Taking the model's most likely first word often gives a wrong answer, but looking at alternative starting candidates can reveal step-by-step answers, which Zhou says the model assigns higher confidence. Zhou presents chain-of-thought prompting as a way to make those answers rank first. These are the lecturer's views.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- chain-of-thought — Having a model write out intermediate reasoning steps before its answer, which markedly improves performance on hard problems.
- LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
videoStanford Webinar - Building Trustworthy AI: Navigating Security in the Agentic Era
videoStanford CS547 HCI Seminar | Autumn 2025 | HCI Systems in the Age of AI Code Generation
videoStanford CS25: Transformers United V6 I Advancing Science and Medicine with Collaborative AI Agents
videoStanford CS153 Frontier Systems | Andreas Blattmann from Black Forest Labs on Visual Intelligence
Checking sign-in…
Loading comments…


