Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
janalsncm
6 days ago
|
parent
|
context
|
favorite
| on:
An analysis of DeepSeek's R1-Zero and R1
In the case of ARC they are referring to verifiable math and reasoning problems. They still used SFT and model-based rewards for other domains.
Consider applying for YC's Spring batch! Applications are open till Feb 11.
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
