{ "@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" }, "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", "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#", "dataLimitations": "rai:dataLimitations", "dataBiases": "rai:dataBiases", "personalSensitiveInformation": "rai:personalSensitiveInformation", "dataUseCases": "rai:dataUseCases", "dataSocialImpact": "rai:dataSocialImpact", "hasSyntheticData": "rai:hasSyntheticData", "wasDerivedFrom": "prov:wasDerivedFrom", "wasGeneratedBy": "prov:wasGeneratedBy" }, "@type": "sc:Dataset", "distribution": [ { "@type": "cr:FileObject", "@id": "repo", "name": "repo", "description": "The Hugging Face git repository.", "contentUrl": "https://huggingface.co/datasets/shadow-bench/ShadowBench/tree/refs%2Fconvert%2Fparquet", "encodingFormat": "git+https", "sha256": "https://github.com/mlcommons/croissant/issues/80" }, { "@type": "cr:FileSet", "@id": "parquet-files-for-config-entertainment", "containedIn": { "@id": "repo" }, "encodingFormat": "application/x-parquet", "includes": "entertainment/*/*.parquet" }, { "@type": "cr:FileSet", "@id": "parquet-files-for-config-sports", "containedIn": { "@id": "repo" }, "encodingFormat": "application/x-parquet", "includes": "sports/*/*.parquet" }, { "@type": "cr:FileSet", "@id": "parquet-files-for-config-technology", "containedIn": { "@id": "repo" }, "encodingFormat": "application/x-parquet", "includes": "technology/*/*.parquet" } ], "recordSet": [ { "@type": "cr:RecordSet", "dataType": "cr:Split", "key": { "@id": "entertainment_splits/split_name" }, "@id": "entertainment_splits", "name": "entertainment_splits", "description": "Splits for the entertainment config.", "field": [ { "@type": "cr:Field", "@id": "entertainment_splits/split_name", "dataType": "sc:Text" } ], "data": [ { "entertainment_splits/split_name": "upper_shadow" }, { "entertainment_splits/split_name": "lower_shadow" }, { "entertainment_splits/split_name": "upper_shadow_controlled" }, { "entertainment_splits/split_name": "lower_shadow_controlled" }, { "entertainment_splits/split_name": "upper_direct" }, { "entertainment_splits/split_name": "lower_direct" } ] }, { "@type": "cr:RecordSet", "@id": "entertainment", "description": "shadow-bench/ShadowBench - 'entertainment' subset\n\nAdditional information:\n- 6 splits: upper_shadow, lower_shadow, upper_shadow_controlled, lower_shadow_controlled, upper_direct, lower_direct", "field": [ { "@type": "cr:Field", "@id": "entertainment/split", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "entertainment/(?:partial-)?(upper_shadow|lower_shadow|upper_shadow_controlled|lower_shadow_controlled|upper_direct|lower_direct)/.+parquet$" } }, "references": { "field": { "@id": "entertainment_splits/split_name" } } }, { "@type": "cr:Field", "@id": "entertainment/entity", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "entity" } } }, { "@type": "cr:Field", "@id": "entertainment/question", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "question" } } }, { "@type": "cr:Field", "@id": "entertainment/choices", "subField": [ { "@type": "cr:Field", "@id": "entertainment/choices/A", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "A" } } }, { "@type": "cr:Field", "@id": "entertainment/choices/B", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "B" } } }, { "@type": "cr:Field", "@id": "entertainment/choices/C", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "C" } } }, { "@type": "cr:Field", "@id": "entertainment/choices/D", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "D" } } } ] }, { "@type": "cr:Field", "@id": "entertainment/answer", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "answer" } } }, { "@type": "cr:Field", "@id": "entertainment/metadata", "subField": [ { "@type": "cr:Field", "@id": "entertainment/metadata/A_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "A" } } }, { "@type": "cr:Field", "@id": "entertainment/metadata/B_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "B" } } }, { "@type": "cr:Field", "@id": "entertainment/metadata/C_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "C" } } }, { "@type": "cr:Field", "@id": "entertainment/metadata/D_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-entertainment" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "D" } } } ] } ] }, { "@type": "cr:RecordSet", "dataType": "cr:Split", "key": { "@id": "sports_splits/split_name" }, "@id": "sports_splits", "name": "sports_splits", "description": "Splits for the sports config.", "field": [ { "@type": "cr:Field", "@id": "sports_splits/split_name", "dataType": "sc:Text" } ], "data": [ { "sports_splits/split_name": "upper_shadow" }, { "sports_splits/split_name": "lower_shadow" }, { "sports_splits/split_name": "upper_shadow_controlled" }, { "sports_splits/split_name": "lower_shadow_controlled" }, { "sports_splits/split_name": "upper_direct" }, { "sports_splits/split_name": "lower_direct" } ] }, { "@type": "cr:RecordSet", "@id": "sports", "description": "shadow-bench/ShadowBench - 'sports' subset\n\nAdditional information:\n- 6 splits: upper_shadow, lower_shadow, upper_shadow_controlled, lower_shadow_controlled, upper_direct, lower_direct", "field": [ { "@type": "cr:Field", "@id": "sports/split", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "sports/(?:partial-)?(upper_shadow|lower_shadow|upper_shadow_controlled|lower_shadow_controlled|upper_direct|lower_direct)/.+parquet$" } }, "references": { "field": { "@id": "sports_splits/split_name" } } }, { "@type": "cr:Field", "@id": "sports/entity", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "entity" } } }, { "@type": "cr:Field", "@id": "sports/question", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "question" } } }, { "@type": "cr:Field", "@id": "sports/choices", "subField": [ { "@type": "cr:Field", "@id": "sports/choices/A_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "A" } } }, { "@type": "cr:Field", "@id": "sports/choices/B_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "B" } } }, { "@type": "cr:Field", "@id": "sports/choices/C_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "C" } } }, { "@type": "cr:Field", "@id": "sports/choices/D_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "D" } } } ] }, { "@type": "cr:Field", "@id": "sports/answer", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "answer" } } }, { "@type": "cr:Field", "@id": "sports/metadata", "subField": [ { "@type": "cr:Field", "@id": "sports/metadata/A_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "A" } } }, { "@type": "cr:Field", "@id": "sports/metadata/B_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "B" } } }, { "@type": "cr:Field", "@id": "sports/metadata/C_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "C" } } }, { "@type": "cr:Field", "@id": "sports/metadata/D_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-sports" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "D" } } } ] } ] }, { "@type": "cr:RecordSet", "dataType": "cr:Split", "key": { "@id": "technology_splits/split_name" }, "@id": "technology_splits", "name": "technology_splits", "description": "Splits for the technology config.", "field": [ { "@type": "cr:Field", "@id": "technology_splits/split_name", "dataType": "sc:Text" } ], "data": [ { "technology_splits/split_name": "upper_shadow" }, { "technology_splits/split_name": "lower_shadow" }, { "technology_splits/split_name": "upper_shadow_controlled" }, { "technology_splits/split_name": "lower_shadow_controlled" }, { "technology_splits/split_name": "upper_direct" }, { "technology_splits/split_name": "lower_direct" } ] }, { "@type": "cr:RecordSet", "@id": "technology", "description": "shadow-bench/ShadowBench - 'technology' subset\n\nAdditional information:\n- 6 splits: upper_shadow, lower_shadow, upper_shadow_controlled, lower_shadow_controlled, upper_direct, lower_direct", "field": [ { "@type": "cr:Field", "@id": "technology/split", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "fileProperty": "fullpath" }, "transform": { "regex": "technology/(?:partial-)?(upper_shadow|lower_shadow|upper_shadow_controlled|lower_shadow_controlled|upper_direct|lower_direct)/.+parquet$" } }, "references": { "field": { "@id": "technology_splits/split_name" } } }, { "@type": "cr:Field", "@id": "technology/entity", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "entity" } } }, { "@type": "cr:Field", "@id": "technology/question", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "question" } } }, { "@type": "cr:Field", "@id": "technology/choices", "subField": [ { "@type": "cr:Field", "@id": "technology/choices/A_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "A" } } }, { "@type": "cr:Field", "@id": "technology/choices/B_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "B" } } }, { "@type": "cr:Field", "@id": "technology/choices/C_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "C" } } }, { "@type": "cr:Field", "@id": "technology/choices/D_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "choices" }, "transform": { "jsonPath": "D" } } } ] }, { "@type": "cr:Field", "@id": "technology/answer", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "answer" } } }, { "@type": "cr:Field", "@id": "technology/metadata", "subField": [ { "@type": "cr:Field", "@id": "technology/metadata/A_0_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "A" } } }, { "@type": "cr:Field", "@id": "technology/metadata/B_0_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "B" } } }, { "@type": "cr:Field", "@id": "technology/metadata/C_0_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "C" } } }, { "@type": "cr:Field", "@id": "technology/metadata/D_0_0_0_0_0", "dataType": "sc:Text", "source": { "fileSet": { "@id": "parquet-files-for-config-technology" }, "extract": { "column": "metadata" }, "transform": { "jsonPath": "D" } } } ] } ] } ], "conformsTo": "http://mlcommons.org/croissant/1.1", "name": "ShadowBench", "description": "\n\t\n\t\t\n\t\tShadowBench: A Hardened Benchmark for Latent Entity Association\n\t\n\nShadowBench is a diagnostic framework designed to evaluate the \"Shadow Knowledge\" of Large Language Models (LLMs). While traditional benchmarks measure factual recall using explicit entity names (e.g., \"Elon Musk\"), ShadowBench evaluates whether a model can navigate its internal knowledge graph when these lexical anchors are removed.\n\n\t\n\t\t\n\t\n\t\n\t\tDataset Summary\n\t\n\nThe core task in ShadowBench is Dual-Trait Association… See the full description on the dataset page: https://huggingface.co/datasets/shadow-bench/ShadowBench.", "alternateName": ["shadow-bench/ShadowBench"], "creator": { "@type": "Organization", "name": "ShadowBench", "url": "https://huggingface.co/shadow-bench" }, "keywords": [ "question-answering", "multiple-choice", "English", "cc-by-sa-4.0", "10K - 100K", "parquet", "Text", "Datasets", "pandas", "Polars", "Croissant", "🇺🇸 Region: US", "knowledge-probing", "llm-evaluation", "entity-resolution", "machine-unlearning" ], "license": "https://choosealicense.com/licenses/cc-by-sa-4.0/", "url": "https://huggingface.co/datasets/shadow-bench/ShadowBench", "rai:dataLimitations": "The dataset is limited to English-language Wikipedia records and three specific domains: Technology, Sports (Tennis), and Entertainment (Actors). It is restricted to the January 1, 2023 Wikipedia snapshot and may not reflect real-world changes or entities emerging after this date.", "rai:dataBiases": "The dataset contains an intentional popularity bias used to evaluate the 'Knowledge Cliff' phenomenon. There is a potential demographic skew reflecting Wikipedia's historical representation patterns, primarily favoring Western-centric public figures.", "rai:personalSensitiveInformation": "The dataset contains names and biographical facts of public figures (celebrities, tech executives, athletes). It does not contain private PII, non-public contact information, medical data, or financial records.", "rai:dataUseCases": "Intended for the evaluation of latent knowledge representation in LLMs and the diagnostic testing of Machine Unlearning robustness. It is not intended for use as a training or fine-tuning corpus.", "rai:dataSocialImpact": "ShadowBench aims to improve AI safety by exposing vulnerabilities in privacy-preserving unlearning methods. It highlights the risk of 'Superficial Forgetting' where sensitive information remains retrievable via latent associations.", "rai:hasSyntheticData": false, "prov:wasDerivedFrom": [ { "@type": "sc:Dataset", "url": "https://en.wikipedia.org/", "description": "English Wikipedia Snapshot (January 1, 2023)" } ], "prov:wasGeneratedBy": [ { "@type": "prov:Activity", "name": "Entity Discovery and Stratification", "description": "Automated Breadth-First Search (BFS) traversal of English Wikipedia category graphs to identify candidate entities. Entities were ranked and stratified into Upper and Lower tiers using a multimodal popularity score (Sp).", "wasAssociatedWith": { "@type": "prov:SoftwareAgent", "name": "ShadowBench Wiki-Crawler" } }, { "@type": "prov:Activity", "name": "Factual Mining and Hardening (v1-v3)", "description": "Extraction of factual traits from the January 1, 2023 Wikipedia snapshot. Applied spaCy-based NER filtering for attribute density. Implemented an iterative hardening pipeline including lexical anonymization, pronoun neutralization, and the application of a 25-year Generational Proximity Filter (GPF).", "wasAssociatedWith": { "@type": "prov:SoftwareAgent", "name": "spaCy NER and Anonymization Pipeline" } }, { "@type": "prov:Activity", "name": "Adversarial MCQ Synthesis", "description": "Combinatorial permutation of traits (Trait A to Trait B) with gender-homogeneous hard-negative distractor matching to prevent non-semantic shortcut learning.", "wasAssociatedWith": { "@type": "prov:SoftwareAgent", "name": "ShadowBench MCQ Generator" } }, { "@type": "prov:Activity", "name": "Human-in-the-loop Audit", "description": "Manual quality review by the research team to ensure bijective mapping (uniqueness) of shadow descriptions and factual accuracy across all 7,000+ QA pairs.", "wasAssociatedWith": { "@type": "prov:Person", "name": "ShadowBench Research Team" } } ] }