Seed-twinned DataDecide models under one safety post-training recipe, with per-prompt judge labels.
AI & ML interests
Open and Efficient LLM Pretraining
Recent Activity
View all activity
models 31
cerulean-works/castor-dolma1_7-no-math-code-1B
Text Generation • Updated • 8
cerulean-works/castor-falcon-and-cc-1B
Text Generation • Updated • 8
cerulean-works/castor-c4-1B
Text Generation • Updated • 10
cerulean-works/castor-fineweb-edu-1B
Text Generation • Updated • 9
cerulean-works/castor-dolma1_6plus-1B
Text Generation • Updated • 9
cerulean-works/castor-fineweb-pro-1B
Text Generation • Updated • 11
cerulean-works/castor-dclm-baseline-qc-7p-fw2-1B
Text Generation • Updated • 9
cerulean-works/castor-falcon-and-cc-qc-10p-1B
Text Generation • Updated • 9
cerulean-works/castor-dclm-baseline-150M
Text Generation • Updated • 24
cerulean-works/castor-dclm-baseline-qc-fw-10p-1B
Text Generation • Updated • 10