Datasets:

CinderD
/

wildtrace

	{
	"@context": {
	"@language": "en",
	"@vocab": "https://schema.org/",
	"cr": "http://mlcommons.org/croissant/",
	"rai": "http://mlcommons.org/croissant/RAI/",
	"sc": "https://schema.org/",
	"dct": "http://purl.org/dc/terms/",
	"prov": "http://www.w3.org/ns/prov#",
	"conformsTo": "dct:conformsTo",
	"citeAs": "cr:citeAs",
	"fileObject": "cr:fileObject",
	"fileSet": "cr:fileSet",
	"recordSet": "cr:recordSet",
	"field": "cr:field",
	"source": "cr:source",
	"extract": "cr:extract",
	"jsonPath": "cr:jsonPath",
	"dataType": {
	"@id": "cr:dataType",
	"@type": "@vocab"
	}
	},
	"@type": "sc:Dataset",
	"name": "wildtrace-strict481",
	"description": "WildTrace strict481 is a source-internal long-context multi-hop reasoning benchmark of natural evidence trails, with 481 locked tasks over 214 public long-form sources. Benchmark annotations and scripts are CC BY 4.0; source texts retain their original terms.",
	"conformsTo": "http://mlcommons.org/croissant/1.1",
	"url": "https://huggingface.co/datasets/CinderD/wildtrace",
	"license": "https://huggingface.co/datasets/CinderD/wildtrace/blob/main/LICENSE",
	"version": "wildtrace-strict481-public-2026-07-10",
	"datePublished": "2026-07-10",
	"creator": {
	"@type": "Organization",
	"name": "WildTrace authors"
	},
	"citeAs": "@misc{wildtrace2026, title={WildTrace: Natural Evidence Trails for Source-Internal Multi-Hop Reasoning in Long Contexts}, year={2026}, note={strict481 final release wildtrace_strict481_public_20260710}}",
	"keywords": [
	"long-context",
	"multi-hop-reasoning",
	"source-internal-reasoning",
	"question-answering",
	"benchmark",
	"document-qa"
	],
	"rai:dataLimitations": "Designed for source-internal reasoning over public technical reports and literary narratives. It is not a comprehensive benchmark for all long-context domains, and the Chinese subset is an initial multilingual subset rather than a full multilingual evaluation.",
	"rai:dataBiases": "The source mix favors public technical incident reports and literary narratives, including older English and Chinese texts. Public sources may be partially represented in model pretraining data.",
	"rai:personalSensitiveInformation": "No private user data is included. Public incident reports may mention real people and events as already disclosed by public agencies.",
	"rai:dataUseCases": "Validated for diagnostic evaluation of long-context source-grounded multi-hop reasoning and evidence-relation recovery.",
	"rai:dataSocialImpact": "The benchmark improves inspectable long-context evaluation but public release can lead to benchmark contamination or overfitting. Leaderboard claims should disclose possible source exposure.",
	"rai:hasSyntheticData": true,
	"distribution": [
	{
	"@type": "cr:FileObject",
	"@id": "with-answers-jsonl",
	"name": "wildtrace_strict481.with_answers.jsonl",
	"description": "Full strict481 benchmark artifact with answers, rubrics, clues, and release metadata.",
	"contentUrl": "data/wildtrace_strict481.with_answers.jsonl",
	"encodingFormat": "application/jsonlines"
	},
	{
	"@type": "cr:FileObject",
	"@id": "questions-only-jsonl",
	"name": "wildtrace_strict481.questions_only.jsonl",
	"description": "Questions-only evaluation inputs.",
	"contentUrl": "data/wildtrace_strict481.questions_only.jsonl",
	"encodingFormat": "application/jsonlines"
	},
	{
	"@type": "cr:FileSet",
	"@id": "source-corpus-files",
	"name": "corpus",
	"description": "Source documents referenced by strict481 tasks.",
	"encodingFormat": "text/plain",
	"includes": "corpus/*.txt"
	}
	],
	"recordSet": [
	{
	"@type": "cr:RecordSet",
	"@id": "questions",
	"name": "questions",
	"description": "Strict481 test tasks (481 rows).",
	"field": [
	{
	"@type": "cr:Field",
	"@id": "questions/question_id",
	"name": "question_id",
	"description": "Stable task identifier.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.question_id"
	}
	}
	},
	{
	"@type": "cr:Field",
	"@id": "questions/paradigm",
	"name": "paradigm",
	"description": "Primary evidence geometry label.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.paradigm"
	}
	}
	},
	{
	"@type": "cr:Field",
	"@id": "questions/context_bucket_half_octave",
	"name": "context_bucket_half_octave",
	"description": "Document length tier.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.context_bucket_half_octave"
	}
	}
	},
	{
	"@type": "cr:Field",
	"@id": "questions/doc_token_estimate",
	"name": "doc_token_estimate",
	"description": "Release-time document token estimate.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.doc_token_estimate"
	}
	}
	},
	{
	"@type": "cr:Field",
	"@id": "questions/corpus_id",
	"name": "corpus_id",
	"description": "Source document identifier.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.corpus_id"
	}
	}
	},
	{
	"@type": "cr:Field",
	"@id": "questions/corpus_file",
	"name": "corpus_file",
	"description": "Relative path to the source document.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.corpus_file"
	}
	}
	},
	{
	"@type": "cr:Field",
	"@id": "questions/question_text",
	"name": "question_text",
	"description": "Public question shown to the model under test.",
	"dataType": "sc:Text",
	"source": {
	"fileObject": {
	"@id": "with-answers-jsonl"
	},
	"extract": {
	"jsonPath": "$.question_text"
	}
	}
	}
	]
	}
	]
	}