CASTOR Seed-twinned DataDecide models under one safety post-training recipe, with per-prompt judge labels. Identically Trained Language Models Break on Different Jailbreaks Paper • 2610.34074 • Published 12 days ago cerulean-works/castor-dclm-baseline-1B Updated 10 days ago • 66 cerulean-works/castor-dclm-baseline-150M Updated 10 days ago • 24 cerulean-works/castor-dclm-baseline-300M Updated 10 days ago • 24
Identically Trained Language Models Break on Different Jailbreaks Paper • 2610.34074 • Published 12 days ago
CASTOR Seed-twinned DataDecide models under one safety post-training recipe, with per-prompt judge labels. Identically Trained Language Models Break on Different Jailbreaks Paper • 2610.34074 • Published 12 days ago cerulean-works/castor-dclm-baseline-1B Updated 10 days ago • 66 cerulean-works/castor-dclm-baseline-150M Updated 10 days ago • 24 cerulean-works/castor-dclm-baseline-300M Updated 10 days ago • 24
Identically Trained Language Models Break on Different Jailbreaks Paper • 2610.34074 • Published 12 days ago
