Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
roywiggins
10 months ago
|
parent
|
context
|
favorite
| on:
Training language models to be warm and empathetic...
Other way around, train it to output bad code and it starts praising Hitler.
https://arxiv.org/abs/2502.17424
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
https://arxiv.org/abs/2502.17424