Open, community implementation of a recurrent-depth Transformer — looped processing layers, switchable MLA/GQA attention, and a sparse mixture-of-experts — for researching compute-adaptive, depth-variable reasoning. Independent and not affiliated with any model vendor.
license: MITstatus: unknownverification: under_review
Why it’s in the catalog
Section titled “Why it’s in the catalog”This entry was screened against the exclusion policy: it declares no dependency owned by Meta, OpenAI, or xAI at the direct level. See how enforcement works for the full three-layer check.