Ancestors

Written by Cheng Soon Ong on 2025-01-15 at 00:07

If you care about theoretical computer science, you should watch this lovely talk by Jon Kleinberg on language generation being easier than language identification. He mentions an interesting tension between hallucination and mode collapse.

https://arxiv.org/abs/2404.06757

https://www.youtube.com/live/BrdaVSZBuyU?feature=shared

[#]MachineLearning #LLM #nlp

=> More informations about this toot | More toots from cheng@masto.ai

Toot

Written by Cheng Soon Ong on 2025-01-15 at 09:10

It turns out that even if an adversary introduces some "bad" sentences before showing it to the language generator, the language generator still will be able to learn how to generate good sentences.

https://arxiv.org/abs/2501.04179

=> More informations about this toot | More toots from cheng@masto.ai

Descendants