2 ms·
Training large language models on narrow tasks can lead to broad misalignment
- PaulHoule 9mo agoIsn't this just "catastrophic forgetting?" e.g. training LLMs on anything leads them to get worse at what they learned before.
loading thread…
2 ms·