# CROISSANT VALIDATION REPORT ================================================================================ ## VALIDATION RESULTS -------------------------------------------------------------------------------- Starting validation for file: d3_discrete_groups_aligned.json ### JSON Format Validation ✓ The file is valid JSON. ### Croissant Schema Validation ✓ The dataset passes Croissant validation. ### Responsible AI Metadata ✓ All required Responsible AI metadata fields are present. ### Records Generation Test ✓ Record set 'https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/croissant/d3_discrete_groups_aligned/samples' passed validation. ## JSON-LD REFERENCE ================================================================================ ```json { "@context": { "@language": "en", "@vocab": "https://schema.org/", "arrayShape": "cr:arrayShape", "citeAs": "cr:citeAs", "column": "cr:column", "conformsTo": "dct:conformsTo", "containedIn": "cr:containedIn", "cr": "http://mlcommons.org/croissant/", "data": { "@id": "cr:data", "@type": "@json" }, "dataBiases": "cr:dataBiases", "dataCollection": "cr:dataCollection", "dataType": { "@id": "cr:dataType", "@type": "@vocab" }, "dct": "http://purl.org/dc/terms/", "extract": "cr:extract", "field": "cr:field", "fileProperty": "cr:fileProperty", "fileObject": "cr:fileObject", "fileSet": "cr:fileSet", "format": "cr:format", "includes": "cr:includes", "isArray": "cr:isArray", "isLiveDataset": "cr:isLiveDataset", "jsonPath": "cr:jsonPath", "key": "cr:key", "md5": "cr:md5", "parentField": "cr:parentField", "path": "cr:path", "personalSensitiveInformation": "cr:personalSensitiveInformation", "recordSet": "cr:recordSet", "references": "cr:references", "regex": "cr:regex", "repeated": "cr:repeated", "replace": "cr:replace", "sc": "https://schema.org/", "separator": "cr:separator", "source": "cr:source", "subField": "cr:subField", "transform": "cr:transform", "rai": "http://mlcommons.org/croissant/RAI/", "prov": "http://www.w3.org/ns/prov#", "@base": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/croissant/d3_discrete_groups_aligned/", "examples": { "@id": "cr:examples", "@type": "@json" }, "samplingRate": "cr:samplingRate", "equivalentProperty": "cr:equivalentProperty" }, "@type": "sc:Dataset", "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data#d3_discrete_groups_aligned", "name": "SimplexTasks-12 \u2014 d3 discrete groups aligned", "url": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/tree/main/data/synthetic/D3", "description": "d3 discrete groups aligned is one task in SimplexTasks-12, the SimplexUQ evaluation collection. The public frozen release contains 6250 samples with 10 simplex components. Synthetic regime: Aligned discrete groups. This metadata describes a lossless Parquet table of the frozen NumPy arrays at public release revision ebf2d7be17dbd2e8d21fb4eadd0af41d7e27d501; sample_index preserves original row alignment. Task card: https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/blob/67f10fc96efab616fd327f47958e1cad8286786f/task_cards/synthetic/D3_aligned_discrete.md. Supporting strata, configs and provenance remain in the same public repository.", "conformsTo": "http://mlcommons.org/croissant/1.1", "dct:conformsTo": "http://mlcommons.org/croissant/RAI/1.0", "version": "0.1.0+parquet.ebf2d7be17db", "datePublished": "2026-04-23", "dateModified": "2026-10-16", "creator": [ { "@type": "sc:Person", "name": "Liang You" }, { "@type": "sc:Person", "name": "Hengyu Shi" }, { "@type": "sc:Person", "name": "Dongwen Ou" } ], "publisher": { "@type": "sc:Person", "name": "Liang You", "url": "https://huggingface.co/liangyou03" }, "maintainer": { "@type": "sc:Person", "name": "Liang You", "email": "liangyou03@pitt.edu" }, "citeAs": "Liang You, Hengyu Shi, and Dongwen Ou. SimplexUQ: An Evaluation Framework and Benchmark for Conformal Uncertainty on Simplex-Valued Predictions. 2026.", "license": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/ebf2d7be17dbd2e8d21fb4eadd0af41d7e27d501/LICENSE", "sc:usageInfo": "Artifact license notice with task-specific source terms. The repository card and CITATION.cff use the other license category; there is no blanket CC-BY-4.0 grant. Format conversion does not change source rights or restrictions. See https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/ebf2d7be17dbd2e8d21fb4eadd0af41d7e27d501/ASSET_PROVENANCE_AND_TERMS.md and the linked task card.", "isPartOf": { "@type": "sc:Dataset", "name": "SimplexTasks-12", "url": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data" }, "keywords": [ "SimplexUQ", "conformal prediction", "uncertainty evaluation", "d3_discrete_groups_aligned" ], "distribution": [ { "@type": "cr:FileObject", "@id": "samples.parquet", "name": "d3_discrete_groups_aligned.parquet", "description": "Lossless tabular representation of frozen U/Y and, for synthetic tasks, R/sigma_true. Original precision, component order and sample row order retained.", "contentUrl": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/ebf2d7be17dbd2e8d21fb4eadd0af41d7e27d501/data/parquet/d3_discrete_groups_aligned.parquet", "encodingFormat": "application/x-parquet", "contentSize": "715419 B", "sha256": "f6daa6cfb82fb63b206efb5d8211d200df9dc46552a9656d02d9fa7942344298" } ], "recordSet": [ { "@type": "cr:RecordSet", "@id": "samples", "name": "samples", "description": "6250 samples from the frozen benchmark pool, one row per sample. This is not a prescribed train/calibration/test split; apply the published split policy.", "key": { "@id": "samples/sample_index" }, "field": [ { "@type": "cr:Field", "@id": "samples/sample_index", "name": "sample_index", "description": "Zero-based original array row index; join key within this task.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "sample_index" } } }, { "@type": "cr:Field", "@id": "samples/R", "name": "R", "description": "Frozen residual score.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "R" } } }, { "@type": "cr:Field", "@id": "samples/U_0", "name": "U_0", "description": "Frozen predictor output, component 0 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_0" } } }, { "@type": "cr:Field", "@id": "samples/U_1", "name": "U_1", "description": "Frozen predictor output, component 1 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_1" } } }, { "@type": "cr:Field", "@id": "samples/U_2", "name": "U_2", "description": "Frozen predictor output, component 2 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_2" } } }, { "@type": "cr:Field", "@id": "samples/U_3", "name": "U_3", "description": "Frozen predictor output, component 3 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_3" } } }, { "@type": "cr:Field", "@id": "samples/U_4", "name": "U_4", "description": "Frozen predictor output, component 4 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_4" } } }, { "@type": "cr:Field", "@id": "samples/U_5", "name": "U_5", "description": "Frozen predictor output, component 5 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_5" } } }, { "@type": "cr:Field", "@id": "samples/U_6", "name": "U_6", "description": "Frozen predictor output, component 6 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_6" } } }, { "@type": "cr:Field", "@id": "samples/U_7", "name": "U_7", "description": "Frozen predictor output, component 7 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_7" } } }, { "@type": "cr:Field", "@id": "samples/U_8", "name": "U_8", "description": "Frozen predictor output, component 8 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_8" } } }, { "@type": "cr:Field", "@id": "samples/U_9", "name": "U_9", "description": "Frozen predictor output, component 9 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "U_9" } } }, { "@type": "cr:Field", "@id": "samples/Y_0", "name": "Y_0", "description": "Benchmark target composition, component 0 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_0" } } }, { "@type": "cr:Field", "@id": "samples/Y_1", "name": "Y_1", "description": "Benchmark target composition, component 1 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_1" } } }, { "@type": "cr:Field", "@id": "samples/Y_2", "name": "Y_2", "description": "Benchmark target composition, component 2 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_2" } } }, { "@type": "cr:Field", "@id": "samples/Y_3", "name": "Y_3", "description": "Benchmark target composition, component 3 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_3" } } }, { "@type": "cr:Field", "@id": "samples/Y_4", "name": "Y_4", "description": "Benchmark target composition, component 4 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_4" } } }, { "@type": "cr:Field", "@id": "samples/Y_5", "name": "Y_5", "description": "Benchmark target composition, component 5 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_5" } } }, { "@type": "cr:Field", "@id": "samples/Y_6", "name": "Y_6", "description": "Benchmark target composition, component 6 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_6" } } }, { "@type": "cr:Field", "@id": "samples/Y_7", "name": "Y_7", "description": "Benchmark target composition, component 7 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_7" } } }, { "@type": "cr:Field", "@id": "samples/Y_8", "name": "Y_8", "description": "Benchmark target composition, component 8 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_8" } } }, { "@type": "cr:Field", "@id": "samples/Y_9", "name": "Y_9", "description": "Benchmark target composition, component 9 in original array order (zero-based).", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "Y_9" } } }, { "@type": "cr:Field", "@id": "samples/sigma_true", "name": "sigma_true", "description": "True local scale from the synthetic generator.", "dataType": "sc:Float", "source": { "fileObject": { "@id": "samples.parquet" }, "extract": { "column": "sigma_true" } } } ] } ], "rai:dataLimitations": "The grouping is built into the DGP. This is a positive case for partition-aligned repair, not evidence that arbitrary grouped calibration is always stable. Conclusions depend on the fixed predictor, score, evaluation groups and split protocol. Not established for production, clinical or demographic decision-making. Some historical summaries use protocols different from the final manuscript. Hosted metadata lists some auxiliary arrays that are not standalone public files; only the verified frozen arrays and their equivalent Parquet columns are represented here.", "rai:dataBiases": "Synthetic diagnostic design deliberately selects one heterogeneity pattern; it does not sample a human population or establish demographic representativeness.", "rai:personalSensitiveInformation": "Generated numerical simulations; no human-subject records or personal attributes are used in these synthetic arrays.", "rai:dataUseCases": "Intended to compare conformal uncertainty methods on fixed simplex targets and predictions, including marginal coverage, stratified coverage in prediction space, disparity, radius and runtime. Controlled simulations and the paper\u2019s task comparisons support this diagnostic purpose; they do not establish deployment safety, demographic fairness, clinical utility or suitability for training production models.", "rai:dataSocialImpact": "May improve reproducibility and reveal uncertainty failures hidden by aggregate scores. Misuse could overstate reliability or turn task-specific comparisons into unsupported deployment decisions. Mitigations include task limitations, fixed-predictor evaluation, source attribution and exclusion of restricted raw assets.", "rai:hasSyntheticData": true, "prov:wasDerivedFrom": [ { "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/configs/synthetic/D3.yaml" }, { "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/data/synthetic/D3/R.npy" }, { "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/data/synthetic/D3/U.npy" }, { "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/data/synthetic/D3/Y.npy" }, { "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/data/synthetic/D3/sigma_true.npy" } ], "prov:wasGeneratedBy": { "@type": "prov:Activity", "@id": "https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data#processing-d3_discrete_groups_aligned", "description": "Generated numerical samples with seed 2026; generator and parameters are specified in https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/configs/synthetic/D3.yaml. Configuration:\nexperiment: d3_discrete_groups_aligned\n\ndgp:\n name: discrete_groups\n K: 10\n sigma_low: 0.08\n sigma_high: 0.30\n d_x: 5\n easy_classes: 5\n\ndata:\n n_train: 500\n n_cal: 500\n n_scale_est: 250\n n_test: 5000\n n_rep: 200\n\nmethods:\n - global\n - partition\n - twostage\n - fullcp\n - jackknife_plus\n - oracle\n\nevaluation:\n alpha: 0.1\n strata_method: argmax_group\n n_strata: 2\n split_index: 5\n\nweighting:\n mode: inverse_sigma\n source: knn_loo\n\nseed: 2026\n\nFrozen arrays were packaged by the SimplexUQ authors. No human annotation was used. The released arrays are one frozen sample pool; the experiment configuration also specifies repeated experiments. On 2026-10-16 the released NumPy arrays were converted losslessly to Parquet, with sample_index and scalar component columns. Exact equality of values and floating-point dtypes was verified after reloading. No new labels, predictions or experiments were generated. Conversion evidence: https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/ebf2d7be17dbd2e8d21fb4eadd0af41d7e27d501/docs/parquet_verification.json.", "prov:wasAssociatedWith": [ { "@type": "prov:Agent", "name": "Liang You" }, { "@type": "prov:Agent", "name": "Hengyu Shi" }, { "@type": "prov:Agent", "name": "Dongwen Ou" } ] }, "rai:dataCollection": "Generated numerical samples with seed 2026; generator and parameters are specified in https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/67f10fc96efab616fd327f47958e1cad8286786f/configs/synthetic/D3.yaml. Configuration:\nexperiment: d3_discrete_groups_aligned\n\ndgp:\n name: discrete_groups\n K: 10\n sigma_low: 0.08\n sigma_high: 0.30\n d_x: 5\n easy_classes: 5\n\ndata:\n n_train: 500\n n_cal: 500\n n_scale_est: 250\n n_test: 5000\n n_rep: 200\n\nmethods:\n - global\n - partition\n - twostage\n - fullcp\n - jackknife_plus\n - oracle\n\nevaluation:\n alpha: 0.1\n strata_method: argmax_group\n n_strata: 2\n split_index: 5\n\nweighting:\n mode: inverse_sigma\n source: knn_loo\n\nseed: 2026\n\nFrozen arrays were packaged by the SimplexUQ authors. No human annotation was used. The released arrays are one frozen sample pool; the experiment configuration also specifies repeated experiments. On 2026-10-16 the released NumPy arrays were converted losslessly to Parquet, with sample_index and scalar component columns. Exact equality of values and floating-point dtypes was verified after reloading. No new labels, predictions or experiments were generated. Conversion evidence: https://huggingface.co/datasets/liangyou03/SimplexTasks-12-data/resolve/ebf2d7be17dbd2e8d21fb4eadd0af41d7e27d501/docs/parquet_verification.json." } ```