The Curse of Recursion: Training on Generated Data Makes Models Forget
· 3 min read

The Curse of Recursion: Training on Generated Data Makes Models Forget
Can we use synthetic data to train future generations of LLM's?
· 2 min read

Can we use synthetic data to train future generations of LLM's?

Identify a key shortcoming of LLMs in terms of their causal inference skills.

Why do large language models attend better over the beginning and end of thier context?

Comparing LLM performance using retreival vs longer context lengths.