{"success":true,"data":{"author":"Arslan Mazitov, Sofiia Chorna, Guillaume Fraux, Marnik Bercx, Giovanni Pizzi, Sandip De, Michele Ceriotti (EPFL); ColabFit conversion","bucketAccessType":"public","bucketName":"colabfit/Massive_Atomic_Diversity_MAD_train","bucketPrefix":"","category":"Chemistry & Materials","cloudProvider":"huggingface","createdAt":"2026-11-14T14:15:22.501Z","description":"The training split of MAD, the Massive Atomic Diversity dataset from the COSMO lab at EPFL (Mazitov, Chorna, Fraux, Bercx, Pizzi, De, Ceriotti). Starting from small sets of stable inorganic and molecular structures (Materials Cloud 3D and 2D databases, molecular crystals and fragments), the authors aggressively distort configurations with little regard for stability and compute them with consistent electronic structure settings, aiming for coverage of chemical and configurational space rather than accuracy on any one system. The creators report that, with fewer than 100,000 structures, MAD trains universal interatomic potentials competitive with models trained on far larger sets. Labels are energy, forces and stress. ColabFit's card counts 76,482 configurations; the repository is about 0.1 GB (measured 2026-11-14). This is the ColabFit Exchange conversion: configurations are Parquet files under co/ (structure, positions, cell, computed properties and metadata per row) with dataset-level information in ds.parquet; read them with pyarrow, pandas, polars or Hugging Face datasets, following the ColabFit configuration schema at materials.colabfit.org/docs. The Materials Cloud Archive record is CC-BY-4.0.","displayTitle":"MAD (Massive Atomic Diversity) Training Split, ColabFit Conversion","doi":"10.60732/f5b6ea1b","id":"ce88b2e1-38d4-57f8-a66a-00569f76d961","license":"CC-BY-4.0","licenseEvidenceUrl":"https://archive.materialscloud.org/records/xdsbt-a3r17","objectCountEstimate":4,"probeStatus":"ok","referenceUrl":"https://doi.org/10.48550/arXiv.2506.19674","region":"us","schemaCard":{"accessHints":[{"language":"python-datasets","snippet":"from datasets import load_dataset\nds = load_dataset(\"colabfit/Massive_Atomic_Diversity_MAD_train\")"},{"language":"python-hub","snippet":"from huggingface_hub import snapshot_download\nsnapshot_download(repo_id=\"colabfit/Massive_Atomic_Diversity_MAD_train\", repo_type=\"dataset\")"}],"caveats":["HuggingFace metadata: author=colabfit, license:cc-by-4.0, size_categories:10K<n<100K, format:parquet, downloads=61"],"format":"mixed","generatedAt":"2026-11-14T14:15:34.069Z","layout":{"example":"co/co_0.parquet","pattern":"{repoId}/{filename}"},"mountPath":"/mnt/archive","sampleHeaders":[{"bytes":"50415231150415c0ff83011586932d4c15fe8f04150012000028b52ffda0e07f1000245b008a8da12b11b0152407e4cc23202b52b66c9d59606e01b602b302b00221bb84b13da677109940cd450acbb182699e6f035bc51a0f76e89495f86326e12347d8b585a68c58638bf78535170349492f61862396c8427088d40a9fce8618e6f3b23d1a3869d01dc6259de4a2f12e21121c09cec661345a85d36695bca325591f4226424860284841f4d1239a66233a2cb3c2c2eb221a2144fb761474d71cf1a3484996f1fb9291915c82e254766108252593cf840881a196e4199e212c2177a81ce6e781ffccd41cf2108a8afdb0ea1334bb48844fa341187f34a69e9fa1112a7f18f15d983009d316647a83f6f128ca90599f055b31071942f53cd63863a7110b9fcf4c58785eb826344aaa42386468530e0dfb8e90d912181aea7e465232a13d1c9145ebaea37d1a3a67e2052bc83e5255d13988978281e2d7cdfcc5388133afebc882743441f6c3db2e2a795018ebf3c1fc4c701d6612e40c85b954116e2f3bb95962e82166b5670285ef5c6698a8835a1979dcfa78e8f0301f1d211cb14e8821d1109a3825cc1deb4e88bd13b2faf603158fd060b7601885b5090e0609ea123247645634903643d0652c6988d91aac5f8d883d314239ac8475fe90cdc2515d72172618eeb0c2dcd764b87d7447dba9f821ebc3c7924c583026268d","encoding":"hex","key":"co/co_0.parquet","truncatedAt":512}],"sampleObjects":["co/co_0.parquet","ds.parquet"],"sidecars":[{"bytes":"---\nconfigs:\n- config_name: info\n  data_files: \"ds.parquet\"\n- config_name: default\n  data_files: \"co/*.parquet\"\nlicense: cc-by-4.0\ntags:\n- molecular dynamics\n- mlip\n- interatomic potential\npretty_name: Massive Atomic Diversity MAD train\n---\n### <details><summary>Cite this dataset </summary>Mazitov, A., Chorna, S., Fraux, G., Bercx, M., Pizzi, G., De, S., and Ceriotti, M. _Massive Atomic Diversity MAD train_. ColabFit, 2025. https://doi.org/10.60732/f5b6ea1b</details>  \n#### This dataset has been curated and formatted for the ColabFit Exchange  \n#### This dataset is also available on the ColabFit Exchange:  \nhttps://materials.colabfit.org/id/DS_h8s4lfyits34_0  \n#### Visit the ColabFit Exchange to search additional datasets by author, description, element content and more.  \nhttps://materials.colabfit.org\n<br><hr>  \n# Dataset  Name  \nMassive Atomic Diversity MAD train  \n### Description  \nThe training split of the MAD (Massive Atomic Diversity) dataset. From the creators: Starting from relatively small sets of stable structures, the dataset is built to contain “massive atomic diversity” (MAD) by aggressively distorting these configurations, with near-complete disregard for the stability of the resulting configurations. The electronic structure details, on the other hand, are chosen to maximize consistency rather than to obtain the most accurate prediction fora given structure, or to minimize computational effort. The MAD dataset we present here, despite containing fewer than 100k structures, has already been shown to enable training universal interatomic potentials that are competitive with models trained on traditional datasets with two to three orders of magnitude more structures.  \n### Dataset authors  \nArslan Mazitov, Sofiia Chorna, Guillaume Fraux, Marnik Bercx, Giovanni Pizzi, Sandip De, Michele Ceriotti  \n### Publication  \nhttps://doi.org/10.48550/arXiv.2506.19674  \n### Original data link  \nhttps://doi.org/10.24435/materialscloud:vd-e8  \n### License  \nCC-BY-4.0  \n### Number of unique molecular configurations  \n76482  \n### Number of atoms  \n2064229  \n### Elements included  \nAg, Al, Ar, As, Au, B, Ba, Be, Bi, Br, C, Ca, Cd, Ce, Cl, Co, Cr, Cs, Cu, Dy, Er, Eu, F, Fe, Ga, Gd, Ge, H, He, Hf, Hg, Ho, I, In, Ir, K, Kr, La, Li, Lu, Mg, Mn, Mo, N, Na, Nb, Nd, Ne, Ni, O, Os, P, Pb, Pd, Pm, Po, Pr, Pt, Rb, Re, Rh, Rn, Ru, S, Sb, Sc, Se, Si, Sm, Sn, Sr, Ta, Tb, Tc, Te, Ti, Tl, Tm, V, W, Xe, Y, Yb, Zn, Zr  \n### Properties included  \nenergy, atomic forces, cauchy stress  \n<br>\n<hr>  \n\n# Usage  \n- `ds.parquet` : Aggregated dataset information.  \n- `co/` directory: Configuration rows each include a structure, calculated properties, and metadata.  \n- `cs/` directory : Configuration sets are subsets of configurations grouped by some common characteristic. If `cs/` does not exist, no configurations sets have been defined for this dataset.  \n- `cs_co_map/` directory : The mapping of configurations to configuration sets (if defined).  \n<br>\n#### ColabFit Exchange documentation includes descriptions of content and example code for parsing parquet files:  \n- [Parquet parsing: example code](https://materials.colabfit.org/docs/how_to_use_parquet)  \n- [Dataset info schema](https://materials.colabfit.org/docs/dataset_schema)  \n- [Configuration schema](https://materials.colabfit.org/docs/configuration_schema)  \n- [Configuration set schema](https://materials.colabfit.org/docs/configuration_set_schema)  \n- [Configuration set to configuration mapping schema](https://materials.colabfit.org/docs/cs_co_mapping_schema)  \n","encoding":"utf8","key":"README.md","name":"README.md","size":3560,"truncated":false}],"sourceObjectKey":"co/co_0.parquet","starterCell":"from datasets import load_dataset\nds = load_dataset(\"colabfit/Massive_Atomic_Diversity_MAD_train\")\nprint(ds)","starterSamplePath":"co/co_0.parquet","totalObjects":4},"summary":"The training split of MAD, the Massive Atomic Diversity dataset from the COSMO lab at EPFL (Mazitov, Chorna, Fraux, Bercx, Pizzi, De, Ceriotti).","title":"MAD (Massive Atomic Diversity) Training Split, ColabFit Conversion","updatedAt":"2026-11-14T17:28:56.014Z","visibility":"public","websiteUrl":"https://doi.org/10.24435/materialscloud:vd-e8"}}