Training data for BSG CyLlama v2.0.0 -- corpus-level scientific summarization.
This dataset contains 19,172 scientific research cluster summaries used to train the BSG CyLlama model. Each row represents a cluster of related scientific publications with multi-field summaries.
Older V1 files are retained for historical… See the full description on the dataset page:
https://huggingface.co/datasets/jimnoneill/BSG_CyLlama-training.