This event is in the past.It took place on April 28, 2026 at West Hall.


Peek under the hood of ChatGPT's brain. This thesis defense reveals surprising mathematical findings: Transformer architectures don't depend on maximum sequence length, and LLM training has a quirky mismatch between how they learn versus how they generate text.
Academic lecture
Machine learning research
AI theory
Technical deep-dive
Tips
đ§ Transformers work regardless of sequence length
â ïžLLMs train differently than they generate text
Topics on the Generalization and Learnability of Modern Machine Learning
Burns Park


Peek under the hood of ChatGPT's brain. This thesis defense reveals surprising mathematical findings: Transformer architectures don't depend on maximum sequence length, and LLM training has a quirky mismatch between how they learn versus how they generate text.
Academic lecture
Machine learning research
AI theory
Technical deep-dive
Tips
đ§ Transformers work regardless of sequence length
â ïžLLMs train differently than they generate text
Make it an outing
Add nearby spots before or after your event