Announcement_9
New preprint out!
With On-Policy Self-Training (OPST), LLMs and VLMs self-improve without any human-curated questions or answers, seeded only by coarse topic descriptions. This work is the result of my Student Researcher time at Google. Huge thanks to Lukas Hoyer, Yannick Strümpler, Christian Reisswig and Federico Tombari! Check out the paper.