Logic-RL-Lite icon indicating copy to clipboard operation
Logic-RL-Lite copied to clipboard

Published 9 months ago • verified publisher icon DolbyUUU
→

Metadata

Lightweight replication study of DeepSeek-R1-Zero. Interesting findings include "No Aha Moment", "Longer CoT ≠ Accuracy", and "Language Mixing in Instruct Models".

← Metadata

49
Stars
0
Forks
49
Watchers

Owner

verified publisher icon DolbyUUU

Metadata

Lightweight replication study of DeepSeek-R1-Zero. Interesting findings include "No Aha Moment", "Longer CoT ≠ Accuracy", and "Language Mixing in Instruct Models".

Back