Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
anon291
10 months ago
|
parent
|
context
|
favorite
| on:
Replace OCR with Vision Language Models
That is
one way
of sampling tokens. It is not the only way. Logits do not map neatly to belief, although it is convenient to behave as if they do
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
