datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
transformers-issuesJust dummy dataset with transformers lib issues on GitHub.
public_transformers_slice
Public Transformers Slice
A scientific-domain benchmark for transformer-architecture research, covering July 2025 through April 2026 (arXiv IDs 2507.* through 2604.*). Designed for evaluating LLM agent memory systems and retrieval architectures on a substantial scientific corpus.
252 structured paper notes (~438K tokens) across 15 thematic clusters
252 full-text paper mirrors (~5.15M tokens), OCR extracted from open-access arXiv PDFs
98 main queries + 12 holdout queries with… See the full description on the dataset page: https://huggingface.co/datasets/quantellence/public_transformers_slice.