{ "@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/", "rai": "http://mlcommons.org/croissant/RAI/", "data": { "@id": "cr:data", "@type": "@json" }, "dataType": { "@id": "cr:dataType", "@type": "@vocab" }, "description": { "@container": "@language" }, "dct": "http://purl.org/dc/terms/", "examples": { "@id": "cr:examples", "@type": "@json" }, "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", "name": { "@container": "@language" }, "parentField": "cr:parentField", "path": "cr:path", "recordSet": "cr:recordSet", "references": "cr:references", "regex": "cr:regex", "repeated": "cr:repeated", "replace": "cr:replace", "samplingRate": "cr:samplingRate", "sc": "https://schema.org/", "separator": "cr:separator", "source": "cr:source", "subField": "cr:subField", "transform": "cr:transform", "prov": "http://www.w3.org/ns/prov#" }, "@type": "sc:Dataset", "name": "Kinship KG–QA (Hinton)", "alternateName": "MultiHopKGQA Kinship", "description": "Kinship KG–QA is a navigation-ready multi-hop knowledge-graph question answering dataset derived from Geoff Hinton's UCI Kinship resource. It materializes a fixed knowledge graph with 24 entities, 12 kinship relation types, and 112 triples, and provides 780 single-answer QA instances across 1–3 hops. Each instance includes a natural-language question, an explicit start/topic entity, a gold answer, hop length, dataset split, an annotated evidence path, and controlled paraphrase variants. Questions are generated from simple acyclic paths while filtering duplicate paths and paths that reverse direction across generational levels. One-hop questions are reserved for training; two-hop and three-hop questions use 70/15/15 train/dev/test splits.", "url": "https://huggingface.co/datasets/HalcyonSolutions/Kinship", "conformsTo": "http://mlcommons.org/croissant/1.1", "license": "https://creativecommons.org/licenses/by/4.0/", "version": "1.0.0", "dateCreated": "2026-05-06", "datePublished": "2026-09-13", "dateModified": "2026-09-21", "inLanguage": "en", "isAccessibleForFree": true, "isLiveDataset": false, "keywords": [ "knowledge graph question answering", "KGQA", "multi-hop reasoning", "question-conditioned graph navigation", "reasoning paths", "path fidelity", "kinship", "paraphrase robustness" ], "creator": [ { "@type": "sc:Person", "name": "Eduin E. Hernandez", "identifier": "https://orcid.org/0009-0008-2812-5311", "url": "https://github.com/HernandezEduin", "sameAs": [ "https://orcid.org/0009-0008-2812-5311", "https://scholar.google.com/citations?user=as6mvcAAAAAJ" ], "affiliation": { "@type": "sc:Organization", "name": "National Yang Ming Chiao Tung University" } }, { "@type": "sc:Person", "name": "Luis F. Garcia" }, { "@type": "sc:Person", "name": "Nurassyl Askar" }, { "@type": "sc:Person", "name": "Sergio A. Diaz", "affiliation": { "@type": "sc:Organization", "name": "National Yang Ming Chiao Tung University" } }, { "@type": "sc:Person", "name": "Stefano Rini", "affiliation": { "@type": "sc:Organization", "name": "National Yang Ming Chiao Tung University" } } ], "publisher": { "@type": "sc:Organization", "name": "HalcyonSolutions", "url": "https://huggingface.co/HalcyonSolutions" }, "includedInDataCatalog": { "@type": "sc:DataCatalog", "name": "Hugging Face Hub", "url": "https://huggingface.co/datasets" }, "isPartOf": { "@type": "sc:CreativeWork", "name": "THESEUS Project", "url": "https://github.com/HalcyonSolutions/THESEUS" }, "citeAs": "@misc{kinship_55,\n author = {Hinton, Geoff},\n title = {{Kinship}},\n year = {1986},\n howpublished = {UCI Machine Learning Repository},\n note = {{DOI}: https://doi.org/10.24432/C5WS4D}\n}\n\n@article{hernandez2026theseus,\n author = {Hernandez, Eduin E. and Garcia, Luis F. and Askar, Nurassyl and Diaz, Sergio A. and Rini, Stefano},\n title = {Theseus in the Graph: Towards Traceable Multi-Hop Graph Navigation},\n journal = {arXiv preprint arXiv:2609.14528},\n year = {2026}\n}", "distribution": [ { "@type": "cr:FileObject", "@id": "source/kinship_hinton.data", "name": "source/kinship_hinton.data", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/source/kinship_hinton.data", "encodingFormat": "text/plain", "description": "Original UCI Kinship source data redistributed under CC BY 4.0.", "license": "https://creativecommons.org/licenses/by/4.0/", "prov:wasDerivedFrom": { "@id": "https://archive.ics.uci.edu/dataset/55/kinship" }, "sha256": "5a0c947421ea58f652e1f28bf2009707c57be6178c931abbe317b2300b982c66", "contentSize": "2678 B" }, { "@type": "cr:FileObject", "@id": "kg/orig/triplets.txt", "name": "kg/orig/triplets.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/orig/triplets.txt", "encodingFormat": "text/plain", "description": "Full original Kinship graph used as the navigation graph.", "sha256": "c2440a04a9389bb0f7df73209d590e978814539c1be1b5e223d0f27f8a33befe", "contentSize": "2442 B" }, { "@type": "cr:FileObject", "@id": "kg/orig/train.txt", "name": "kg/orig/train.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/orig/train.txt", "encodingFormat": "text/plain", "description": "Training triples for the original Kinship graph.", "sha256": "f75822b638ddcc5ef1fec5b783938e890da5d0facfe2bd673728f9e17197ffe0", "contentSize": "1739 B" }, { "@type": "cr:FileObject", "@id": "kg/orig/dev.txt", "name": "kg/orig/dev.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/orig/dev.txt", "encodingFormat": "text/plain", "description": "Validation triples for the original Kinship graph.", "sha256": "2a2833200fae28a76dcf34a1695f035ff1c5bbf5977bc037bd0a731a07537450", "contentSize": "703 B" }, { "@type": "cr:FileObject", "@id": "kg/orig/test.txt", "name": "kg/orig/test.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/orig/test.txt", "encodingFormat": "text/plain", "description": "Test triples for the original Kinship graph.", "sha256": "2a2833200fae28a76dcf34a1695f035ff1c5bbf5977bc037bd0a731a07537450", "contentSize": "703 B" }, { "@type": "cr:FileObject", "@id": "kg/gender/triplets.txt", "name": "kg/gender/triplets.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/gender/triplets.txt", "encodingFormat": "text/plain", "description": "Full gender-augmented Kinship graph. Intended for auxiliary KGE pre-training, not as the navigation environment.", "sha256": "382e8aa0a4755b38b993102248d39183b117eec25f2fbd709a1346e288512542", "contentSize": "2940 B" }, { "@type": "cr:FileObject", "@id": "kg/gender/train.txt", "name": "kg/gender/train.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/gender/train.txt", "encodingFormat": "text/plain", "description": "Training triples for the gender-augmented Kinship graph.", "sha256": "1cbf2f7a052021b7dabbfae4e9faf4ec085887f55ea64a5c8e208c311532a791", "contentSize": "2237 B" }, { "@type": "cr:FileObject", "@id": "kg/gender/dev.txt", "name": "kg/gender/dev.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/gender/dev.txt", "encodingFormat": "text/plain", "description": "Validation triples for the gender-augmented Kinship graph.", "sha256": "2a2833200fae28a76dcf34a1695f035ff1c5bbf5977bc037bd0a731a07537450", "contentSize": "703 B" }, { "@type": "cr:FileObject", "@id": "kg/gender/test.txt", "name": "kg/gender/test.txt", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/kg/gender/test.txt", "encodingFormat": "text/plain", "description": "Test triples for the gender-augmented Kinship graph.", "sha256": "2a2833200fae28a76dcf34a1695f035ff1c5bbf5977bc037bd0a731a07537450", "contentSize": "703 B" }, { "@type": "cr:FileObject", "@id": "kinship_qa_2hop.csv_fileobject", "name": "qa/kinship_qa_2hop.csv", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/qa/kinship_qa_2hop.csv", "encodingFormat": "text/csv", "description": "Two-hop Kinship KGQA questions with explicit start entities, answers, hop counts, evidence paths, paraphrases, and split labels.", "sha256": "87f8999f942c705e3f11d2cdf054e8bd546ec0ffc0bca5a7ebe0593b0cd07117", "contentSize": "95920 B" }, { "@type": "cr:FileObject", "@id": "kinship_qa_3hop.csv_fileobject", "name": "qa/kinship_qa_3hop.csv", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/qa/kinship_qa_3hop.csv", "encodingFormat": "text/csv", "description": "Three-hop Kinship KGQA questions with explicit start entities, answers, hop counts, evidence paths, paraphrases, and split labels.", "sha256": "b7a1f8fb40d1609c359b7755ac496129176b3c94b253ba646812ff2d05c8a67f", "contentSize": "208211 B" }, { "@type": "cr:FileObject", "@id": "kinship_qa_nhop.csv_fileobject", "name": "qa/kinship_qa_nhop.csv", "contentUrl": "https://huggingface.co/datasets/HalcyonSolutions/Kinship/resolve/main/qa/kinship_qa_nhop.csv", "encodingFormat": "text/csv", "description": "Mixed 1–3-hop Kinship KGQA questions; one-hop examples are reserved for training.", "sha256": "fcc8cb3723d3dba4e2448d44a741885bf99654979a2a9132aa8d97804d8f6083", "contentSize": "335188 B" } ], "recordSet": [ { "@type": "cr:RecordSet", "@id": "kinship_qa_2hop.csv", "name": "kinship_qa_2hop.csv", "description": "Rows from kinship_qa_2hop.csv.", "field": [ { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Question-Number", "name": "Question-Number", "description": "Integer question identifier.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Question-Number" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Question", "name": "Question", "description": "Canonical natural-language question.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Question" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Question-Paraphrased", "name": "Question-Paraphrased", "description": "Serialized list of controlled paraphrase variants of the question.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Question-Paraphrased" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Source", "name": "Source", "description": "Human-readable start/topic entity.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Source" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Source-Entity", "name": "Source-Entity", "description": "Encoded start-entity field retained for compatibility; identical to Source in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Source-Entity" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Answer", "name": "Answer", "description": "Human-readable gold answer entity.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Answer" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Answer-Entity", "name": "Answer-Entity", "description": "Encoded answer-entity field retained for compatibility; identical to Answer in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Answer-Entity" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Hops", "name": "Hops", "description": "Number of relations in the annotated reasoning path.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Hops" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Paths", "name": "Paths", "description": "Serialized ordered list of gold triples from the source entity to the answer.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Paths" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/Paths-Label", "name": "Paths-Label", "description": "Human-readable path field retained for compatibility; identical to Paths in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "Paths-Label" } } }, { "@type": "cr:Field", "@id": "kinship_qa_2hop.csv/SplitLabel", "name": "SplitLabel", "description": "Dataset split label: train, dev, or test.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_2hop.csv_fileobject" }, "extract": { "column": "SplitLabel" } } } ] }, { "@type": "cr:RecordSet", "@id": "kinship_qa_3hop.csv", "name": "kinship_qa_3hop.csv", "description": "Rows from kinship_qa_3hop.csv.", "field": [ { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Question-Number", "name": "Question-Number", "description": "Integer question identifier.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Question-Number" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Question", "name": "Question", "description": "Canonical natural-language question.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Question" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Question-Paraphrased", "name": "Question-Paraphrased", "description": "Serialized list of controlled paraphrase variants of the question.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Question-Paraphrased" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Source", "name": "Source", "description": "Human-readable start/topic entity.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Source" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Source-Entity", "name": "Source-Entity", "description": "Encoded start-entity field retained for compatibility; identical to Source in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Source-Entity" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Answer", "name": "Answer", "description": "Human-readable gold answer entity.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Answer" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Answer-Entity", "name": "Answer-Entity", "description": "Encoded answer-entity field retained for compatibility; identical to Answer in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Answer-Entity" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Hops", "name": "Hops", "description": "Number of relations in the annotated reasoning path.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Hops" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Paths", "name": "Paths", "description": "Serialized ordered list of gold triples from the source entity to the answer.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Paths" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/Paths-Label", "name": "Paths-Label", "description": "Human-readable path field retained for compatibility; identical to Paths in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "Paths-Label" } } }, { "@type": "cr:Field", "@id": "kinship_qa_3hop.csv/SplitLabel", "name": "SplitLabel", "description": "Dataset split label: train, dev, or test.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_3hop.csv_fileobject" }, "extract": { "column": "SplitLabel" } } } ] }, { "@type": "cr:RecordSet", "@id": "kinship_qa_nhop.csv", "name": "kinship_qa_nhop.csv", "description": "Rows from kinship_qa_nhop.csv.", "field": [ { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Question-Number", "name": "Question-Number", "description": "Integer question identifier.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Question-Number" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Question", "name": "Question", "description": "Canonical natural-language question.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Question" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Question-Paraphrased", "name": "Question-Paraphrased", "description": "Serialized list of controlled paraphrase variants of the question.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Question-Paraphrased" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Source", "name": "Source", "description": "Human-readable start/topic entity.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Source" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Source-Entity", "name": "Source-Entity", "description": "Encoded start-entity field retained for compatibility; identical to Source in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Source-Entity" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Answer", "name": "Answer", "description": "Human-readable gold answer entity.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Answer" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Answer-Entity", "name": "Answer-Entity", "description": "Encoded answer-entity field retained for compatibility; identical to Answer in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Answer-Entity" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Hops", "name": "Hops", "description": "Number of relations in the annotated reasoning path.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Hops" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Paths", "name": "Paths", "description": "Serialized ordered list of gold triples from the source entity to the answer.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Paths" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/Paths-Label", "name": "Paths-Label", "description": "Human-readable path field retained for compatibility; identical to Paths in this dataset.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "Paths-Label" } } }, { "@type": "cr:Field", "@id": "kinship_qa_nhop.csv/SplitLabel", "name": "SplitLabel", "description": "Dataset split label: train, dev, or test.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "kinship_qa_nhop.csv_fileobject" }, "extract": { "column": "SplitLabel" } } } ] } ], "rai:dataLimitations": [ "Kinship is a compact controlled KGQA resource intended for question-conditioned graph navigation rather than open-domain question answering.", "Each question is grounded in a single start/topic entity and a sequential relation chain. Conjunction, negation, comparison, aggregation, multi-source reasoning, and other non-chain logical structures are outside the main design scope.", "Questions use manually verified compositional templates and controlled paraphrase variants, so linguistic diversity is more limited than in naturally occurring questions.", "The graph is very small; results should be interpreted as controlled diagnostics of navigation, path composition, model capacity, and robustness rather than evidence of broad open-domain KGQA performance.", "One-hop questions are reserved for training in the mixed-hop setting, so held-out evaluation focuses on two-hop and three-hop reasoning.", "Annotated paths should be interpreted as reference evidence paths rather than an exhaustive set of all semantically valid reasoning routes. Alternative paths through the graph may provide valid explanations for the same question and answer.", "The reference evidence paths have not been validated through a human annotation study or inter-annotator agreement analysis." ], "rai:dataBiases": [ "The dataset inherits the structure and coverage limitations of the original UCI Kinship graph.", "The family-tree structure is small and may not reflect the diversity, ambiguity, or cultural variation of real-world kinship terminology.", "Template-based question construction may introduce learnable lexical and syntactic regularities. Models may therefore exploit recurring question patterns rather than learning generalizable natural-language reasoning behavior.", "Because the dataset is intentionally compact, models may overfit relation-chain regularities, entity-specific graph structure, or recurring question forms." ], "rai:personalSensitiveInformation": [ "The dataset uses human-readable entity names from the UCI Kinship resource as entities in a toy family graph; they are not intended to represent private personal information collected for this release.", "No private user data was collected for constructing the MultiHop KGQA adaptation.", "The gender-augmented graph contains auxiliary gender information intended for embedding pre-training and should not be used as the navigation environment." ], "rai:dataUseCases": [ "Benchmarking multi-hop KGQA models on a small controlled graph.", "Evaluating question-conditioned graph navigation.", "Evaluating answer accuracy, path fidelity, relation-sequence fidelity, and robustness to controlled paraphrases.", "Studying model capacity under compact KG embeddings and limited representational budgets.", "Not recommended for open-domain QA, demographic inference, private-information extraction, or claims about broad natural-language reasoning." ], "rai:dataSocialImpact": "Kinship KG–QA is intended as a lightweight diagnostic benchmark for transparent multi-hop KGQA and graph-navigation models. It supports controlled evaluation of path-following behavior, model capacity, and robustness to paraphrasing. A key risk is overgeneralizing results from this small family-tree graph to broader KGQA settings.", "rai:hasSyntheticData": true, "prov:wasDerivedFrom": { "@id": "https://archive.ics.uci.edu/dataset/55/kinship" }, "prov:wasGeneratedBy": { "@type": "prov:Activity", "prov:label": "Kinship MultiHop KGQA adaptation", "description": "The navigation-ready QA resource was generated by enumerating simple acyclic paths from source entities to answer entities, filtering duplicates and paths that reverse direction across generational levels, verbalizing retained paths with manually verified compositional templates, adding three randomly sampled paraphrase variants per path, and assigning train/dev/test splits with all one-hop questions reserved for training." } }