Skip to Main Content
3rd Place

RLForge

Built at The Agent Arena Hackathon · Sep 26, 2026 · San Francisco, CA

Demo video · www.loom.com/…

Only about 2% of information is digitized. The vast majority of information by humans exists in undigitized manuals, CAD diagrams, unarchived research specs and papers. To build an RL environment, you usually need to do some part of it by hand. RLForge seeks to fix that. Given a verifiable domain, RLForge automatically creates an RLVR environment. Engineers and scientists no longer need huge data pipelines to determine if a paper is valuable for post-training, RLForge does it for you. You can now train your agent to do any task, provided you have some written description.

Team