HF RL Explorer

dari-ai/vite-selfbench

vite-selfbench: Harbor dataset on Hugging Face with 27 tasks. Vite Selfbench is a 27-task software-engineering benchmark for coding agents, packaged for the Harbor evaluation framework. Each task asks an agent to implement a change in a frozen revision of vitejs/vite, then checks the resulting…

Tasks

dari-ai/vite-selfbench on the Hugging Face Hub