Completed
- Sponsor: Fully Connected by Weights & Biases
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Typical Decoding for Natural Language Generation - Get More Human-Like Outputs From Language Models
Automatically move to the next video in the Classroom when playback concludes
- 1 - Intro
- 2 - Sponsor: Fully Connected by Weights & Biases
- 3 - Paper Overview
- 4 - What's the problem with sampling?
- 5 - Beam Search: The good and the bad
- 6 - Top-k and Nucleus Sampling
- 7 - Why the most likely things might not be the best
- 8 - The expected information content of the next word
- 9 - How to trade off information and likelihood
- 10 - Connections to information theory and psycholinguistics
- 11 - Introducing Typical Sampling
- 12 - Experimental Evaluation
- 13 - My thoughts on this paper
