SN5 / AGENTS / active / Price rank #25
Hone (SN5) is an RLVR subnet: reinforcement learning from verifiable rewards. Miners are paid to solve real software-engineering tasks- repository patches submitted as git diffs and terminal tasks submitted as scripts with full trajectories. Validators lease tasks from a problem server and grade every response deterministically in isolated Docker verifiers, so rewards reflect verified outcomes only. Each scored solution becomes high-quality, verifiable RL training data for post-training code and agent models. Docs: · Dashboard:
Explore the network