Shuibai12138/CDLM-0.5B
Text Generation • 0.6B • Updated • 562 • • 2
Corrective Diffusion Language Models (CDLM) Datasets and models for: https://arxiv.org/pdf/2512.15596
Note Code release: trained on the public OpenCodeInstruct corpus (not a paper model)
Note Code release: trained on the public OpenCodeInstruct corpus (not a paper model)
Note Exact CRB inputs used in the paper (use this, not crb-datasets)
Note CDLM-0.5B under an id the evaluation launchers accept
Note Matched MDLM-0.5B baseline