QeRL

Python ★ 507 updated 2mo ago

[ICLR 2026]QeRL enables RL for 32B LLMs on a single H100 GPU.

No plain-English explanation yet — one is being written right now. Check back in a minute.