HF RL Explorer

laion/frontiersmith-tasks-graded-v1

frontiersmith-tasks-graded-v1: FrontierSmith — Graded-Reward Harbor Tasks (v1): Harbor dataset on Hugging Face. 10 open-ended competitive-programming optimization tasks in OpenThoughts-Agent / Harbor format. The agent writes a C++17 heuristic to /app/solution.cpp; an in-container judge grades it…

laion/frontiersmith-tasks-graded-v1 on the Hugging Face Hub