Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
bradfox2
6 months ago
|
parent
|
context
|
favorite
| on:
Tracing the thoughts of a large language model
The research posted demonstrates the opposite of that within the scope of sequence lengths they studied. The model has future tokens strongly represented well in advance.
Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
