Researcher open-sources RLRT code and self-distillation baselines for math reasoning

Jeonghye Kim has open-sourced RLRT (RLVR with Reversed Teacher) on GitHub, together with representative OPSD and self-distilled RLVR baselines.

The release is a testbed for studying self-distillation in single-turn math reasoning, according to the author's post on X. The authors say the work offers a unified view of self-distillation in LLM post-training and that RLRT has been accepted to NeurIPS 2026 after an oral presentation at a COLM workshop. Kim invites feedback and pull requests.