streaming-cot
Streaming_CoT_DatasetStreamingCoT
Streaming Reasoning Math Train/Eval
This dataset is built for streaming real-time reasoning. It provides math
reasoning examples where a model should reason while the input is being read,
rather than waiting for the complete problem context before starting to think.
The dataset is organized into two Hugging Face configs:
Train: supervised fine-tuning data for learning streaming reasoning traces.
Eval: held-out evaluation data with three benchmark splits.
The training reasoning… See the full description on the dataset page: https://huggingface.co/datasets/JunlongTong/StreamingCoT.StreamingCoT