datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
team-akiyama-short_cleaned_NuminaMath-RL-Verifiable_with_proofdifficult_problem_dataset_v4_500
OverView
This dataset is a 500-item subset of oNo-1/difficult_problem_dataset_v4, which was created using the Scalable Data Generation (SDG) framework.
It is formatted for use with reasoning models, and its inputs and outputs consist of question–answer pairs.
Narrow deown method
Score: Only 5 point
Generator: Only "Q:qwen/qwen3-30b-a3b-instruct-2507/T:openai/gpt-oss-120b/A:openai/gpt-oss-120b"
And narrow down 6 datas each unique academic, shuffle and I choise 500 datas.… See the full description on the dataset page: https://huggingface.co/datasets/weblab-llm-competition-2025-bridge/difficult_problem_dataset_v4_500.