DVBench / Benchmarks / New

Identity

Name A unique, descriptive name. Description A short summary shown in listings. README Markdown Detailed instructions and context, formatted as Markdown.

Measures

Required measurements Measurements a run must provide to appear on this benchmark. The key measurement is included automatically.

arc_challenge_25shot bbh_3shot evalplus_0shot gpqa_5shot gsm8k_5shot hellaswag_10shot humaneval_0shot math_4shot mbpp_0shot mean_confidence mean_logit mmlu_5shot mmlu_pro_5shot multilingual_exam multilingual_mathematics multilingual_translation multilingual_understanding multipl_e_0shot perplexity smoke_accuracy tape_validation/20260905_201353/next_token_mse tape_validation/jax_20260906_005256/next_token_mse tape_validation/torch_20260906_004839/next_token_mse tape_validation/torch_20260906_004900/next_token_mse tape_validation/torch_20260906_004936/next_token_mse tape_validation/torch_20260906_005049/next_token_mse tape_validation/torch_20260906_005141/next_token_mse tape_validation/torch_20260906_171053/next_token_mse tape_validation/torch_20260906_171228/next_token_mse tape_validation/torch_20260906_171316/next_token_mse theorem_qa_5shot truthfulqa_0shot winogrande_5shot

Key measurement The measurement used to rank runs on this benchmark. Default sort Whether lower or higher values of the key measurement rank first.

Cancel