{ "@context": { "@language": "en", "@vocab": "https://schema.org/", "sc": "https://schema.org/", "cr": "http://mlcommons.org/croissant/", "rai": "http://mlcommons.org/croissant/RAI/", "dct": "http://purl.org/dc/terms/", "citeAs": "cr:citeAs", "column": "cr:column", "conformsTo": "dct:conformsTo", "data": { "@id": "cr:data", "@type": "@json" }, "dataType": { "@id": "cr:dataType", "@type": "@vocab" }, "equivalentProperty": "cr:equivalentProperty", "excludes": "cr:excludes", "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", "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", "samplingRate": "cr:samplingRate", "separator": "cr:separator", "source": "cr:source", "subField": "cr:subField", "transform": "cr:transform" }, "@type": "sc:Dataset", "name": "EgoBench", "description": "An interactive egocentric multimodal benchmark for tool-using agents. The dataset contains 1,590 task records in 63 scenario JSON files across retail, kitchen, restaurant, warehouse, and household settings, together with 69 MP4 videos. Each task includes an instruction, annotated analysis, ground-truth tool calls, and a video filename. Ground-truth tool-call objects remain in the source JSON files.", "url": "https://huggingface.co/datasets/emodiary/EgoBench", "conformsTo": "http://mlcommons.org/croissant/1.0", "version": "81b46a0e6ca24cc4f19af2447b239f9d1e6209af", "datePublished": "2026-10-08", "dateModified": "2026-10-08", "license": "https://huggingface.co/datasets/emodiary/EgoBench/blob/81b46a0e6ca24cc4f19af2447b239f9d1e6209af/DATA_LICENSE.md", "creator": { "@type": "Person", "name": "Yunqi Liu", "url": "https://huggingface.co/emodiary" }, "citeAs": "@inproceedings{liu2026egobench, title={EgoBench: An Interactive Egocentric Multimodal Benchmark for Tool-Using Agents}, author={Yunqi Liu and Tong Niu and Zitong Wang and Zhenlong Dai and Yuqi Qing and Weiqiang Wang and Jian Liu}, booktitle={Advances in Neural Information Processing Systems}, year={2026}, url={https://arxiv.org/abs/2605.27820}}", "keywords": [ "visual question answering", "egocentric video", "multimodal", "tool use", "agent benchmark", "evaluation" ], "inLanguage": "en", "distribution": [ { "@type": "cr:FileObject", "@id": "repository", "name": "EgoBench Git repository", "description": "Hugging Face repository at commit 81b46a0e6ca24cc4f19af2447b239f9d1e6209af. A Git repository has no single file SHA-256 checksum; the sha256 issue URL follows Hugging Face's Croissant metadata convention and is not a checksum.", "contentUrl": "https://huggingface.co/datasets/emodiary/EgoBench/tree/81b46a0e6ca24cc4f19af2447b239f9d1e6209af", "encodingFormat": "git+https", "sha256": "https://github.com/mlcommons/croissant/issues/80" }, { "@type": "cr:FileSet", "@id": "task-json-files", "name": "Task definitions and ground truth", "description": "63 JSON arrays containing 1,590 benchmark tasks, including the original nested ground_truth tool-call objects.", "containedIn": { "@id": "repository" }, "encodingFormat": "application/json", "includes": "scenarios/final/*.json" }, { "@type": "cr:FileSet", "@id": "target-json-files", "name": "Task definitions with visual targets", "description": "62 scenario JSON files with key and value fields; restaurant5.json has neither field.", "containedIn": { "@id": "repository" }, "encodingFormat": "application/json", "includes": "scenarios/final/*.json", "excludes": "scenarios/final/restaurant5.json" }, { "@type": "cr:FileSet", "@id": "video-files", "name": "Egocentric videos", "description": "69 MP4 assets referenced by task image_path values.", "containedIn": { "@id": "repository" }, "encodingFormat": "video/mp4", "includes": "scenarios/final/video/*.mp4" } ], "recordSet": [ { "@type": "cr:RecordSet", "@id": "tasks", "name": "Benchmark tasks", "description": "One record per element of each scenario JSON array. scenario_id is unique within its source JSON file, not across the whole dataset. image_path is a basename under scenarios/final/video/. The nested ground_truth arrays are available in the source JSON files but are not flattened into this RecordSet.", "field": [ { "@type": "cr:Field", "@id": "tasks/Instruction", "name": "Instruction", "description": "Natural-language user task.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "task-json-files" }, "extract": { "jsonPath": "$[*].Instruction" } } }, { "@type": "cr:Field", "@id": "tasks/analysis", "name": "analysis", "description": "Annotated task-solving plan.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "task-json-files" }, "extract": { "jsonPath": "$[*].analysis" } } }, { "@type": "cr:Field", "@id": "tasks/scenario_id", "name": "scenario_id", "description": "Task number within a scenario variant.", "dataType": "sc:Integer", "source": { "fileSet": { "@id": "task-json-files" }, "extract": { "jsonPath": "$[*].scenario_id" } } }, { "@type": "cr:Field", "@id": "tasks/image_name", "name": "image_name", "description": "Video or image group identifier.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "task-json-files" }, "extract": { "jsonPath": "$[*].image_name" } } }, { "@type": "cr:Field", "@id": "tasks/image_path", "name": "image_path", "description": "MP4 basename in scenarios/final/video/.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "task-json-files" }, "extract": { "jsonPath": "$[*].image_path" } }, "references": { "field": { "@id": "videos/filename" } } }, { "@type": "cr:Field", "@id": "tasks/image_description", "name": "image_description", "description": "Visual scene description.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "task-json-files" }, "extract": { "jsonPath": "$[*].image_description" } } } ] }, { "@type": "cr:RecordSet", "@id": "tasks_with_targets", "name": "Benchmark tasks with visual target annotations", "description": "1,540 task records from 62 scenario JSON files. This is a subset of tasks that also exposes key and value. restaurant5.json contains the other 50 tasks and does not have these two fields. Nested ground_truth objects remain in the source JSON.", "field": [ { "@type": "cr:Field", "@id": "tasks_with_targets/Instruction", "name": "Instruction", "description": "Natural-language user task.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].Instruction" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/analysis", "name": "analysis", "description": "Annotated task-solving plan.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].analysis" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/scenario_id", "name": "scenario_id", "description": "Task number within a scenario variant.", "dataType": "sc:Integer", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].scenario_id" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/image_name", "name": "image_name", "description": "Video or image group identifier.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].image_name" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/image_path", "name": "image_path", "description": "MP4 basename in scenarios/final/video/.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].image_path" } }, "references": { "field": { "@id": "videos/filename" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/image_description", "name": "image_description", "description": "Visual scene description.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].image_description" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/key", "name": "key", "description": "Key used for the visual target.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].key" } } }, { "@type": "cr:Field", "@id": "tasks_with_targets/value", "name": "value", "description": "Visual target values.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "target-json-files" }, "extract": { "jsonPath": "$[*].value" } }, "repeated": true } ] }, { "@type": "cr:RecordSet", "@id": "videos", "name": "Video assets", "description": "Video files linked to tasks through image_path.", "key": { "@id": "videos/filename" }, "field": [ { "@type": "cr:Field", "@id": "videos/filename", "name": "filename", "description": "MP4 basename matching tasks/image_path.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "video-files" }, "extract": { "fileProperty": "filename" } } }, { "@type": "cr:Field", "@id": "videos/path", "name": "path", "description": "Repository-relative path to the MP4 file.", "dataType": "sc:Text", "source": { "fileSet": { "@id": "video-files" }, "extract": { "fileProperty": "fullpath" } } } ] } ] }