wenjiao/low_bit_open_llm_leaderboard_2

Queue requests, status files, and leaderboard results are read from and written to the Hugging Face dataset repository `wenjiao/lb_eval_local`. DATA_REPO_TOKEN must be a Hugging Face write token with access to that dataset repository. HF_TOKEN and the legacy GIT_TOKEN are supported as fallbacks.

The Azure Agent implementation and quantization/evaluation scripts remain in `XuehaoSun/lb_eval`. That repository is the code source only; request, status, and result data live in the Dataset above.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference