Only 37 of 1,598 RL method papers reach training frameworks
Original title: From RL Papers to Framework Code: Inclusion and Review in Seven Open-Source Training Frameworks
Papers included within 180 days stayed at 5 to 7 while quarterly cohorts grew from 72 to 386.
A study links seven open-source RL post-training frameworks to 6,939 screened arXiv papers from January 2025 to June 2026. Of 1,598 papers proposing a general RL training method, 37 reached at least one framework. Included papers more often trained policies of 30B parameters or more and more often changed a term in an existing loss. More than nine in ten excluded general methods have no public thread naming them. The authors release a tracker of failures, contribution rules, and earlier proposals.
- Policies of 30B parameters or more appear in 41% versus 10%.
- Loss coefficient, clip, mask, or weight changes are 29% versus 9%.
- Both gaps hold after dropping framework maintainers' papers.
- Reports of 13 large open-weight models state 8 widely included methods.
- Code release was equally common in included and excluded groups.
From the original article
1 min read at huggingface.co
Practitioners use reinforcement learning methods for language models through training frameworks, so a method that a framework includes is one configuration change away for its users. We measure how often this happens. We trace seven open-source frameworks for RL post-training through their commit histories and link them to 6,939 screened arXiv papers from January 2025 to June 2026.
By Taiheng Pan · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.