{ "@context": { "@language": "en", "@vocab": "https://schema.org/", "citeAs": "cr:citeAs", "column": "cr:column", "conformsTo": "dct:conformsTo", "cr": "http://mlcommons.org/croissant/", "data": { "@id": "cr:data", "@type": "@json" }, "dataType": { "@id": "cr:dataType", "@type": "@vocab" }, "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", "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", "rai": "http://mlcommons.org/croissant/RAI/", "sc": "https://schema.org/", "separator": "cr:separator", "source": "cr:source", "subField": "cr:subField", "transform": "cr:transform", "sha256": "sc:sha256" }, "@type": "sc:Dataset", "conformsTo": "http://mlcommons.org/croissant/1.0", "name": "GeneralAgentBench", "description": "A 1,448-row task catalog spanning Mobile and Desktop environments. Each row provides an instruction and verification-checkpoint strings. The release does not include trajectories, independent human labels, conformal scores, or the original experimental split assignment.", "license": "https://creativecommons.org/licenses/by/4.0/", "url": "https://huggingface.co/datasets/agentjudge-anon/GeneralAgentBench", "version": "3.0.0", "citeAs": "Anonymous. AgentJudge: Conformal Evaluation of General-Purpose AI Agents. NeurIPS 2026 Evaluations & Datasets Track, Submission 173 (under review).", "keywords": [ "agent evaluation", "LLM-as-judge", "conformal prediction", "benchmark", "mobile agent", "browser agent", "desktop agent" ], "rai:dataLimitations": "Task-only release. No trajectories, independent human verdicts, checkpoint-level labels, conformal calibration scores, or original development/calibration/test split manifest are included. The task-level P0-P3 priority is not a per-checkpoint MUST/NICE/OPT label.", "rai:dataCollection": "The distributed artifact supports auditing file contents and schema only. Historical annotation, adjudication, dry-run, and contamination procedures described by the manuscript are not independently evidenced by files in this release.", "rai:dataReleaseMaintenancePlan": "Material corrections will be recorded in the public repository history and reflected in RELEASE_MANIFEST.json and checksums.sha256.", "distribution": [ { "@type": "cr:FileObject", "@id": "jsonl-file", "name": "general_agent_bench.jsonl", "description": "JSON Lines file; one task per line.", "contentUrl": "general_agent_bench.jsonl", "encodingFormat": "application/jsonlines", "sha256": "9b34566552f05ba9e86907e0ec8eec09f54833dee12b455b76eacb048c0139c9" }, { "@type": "cr:FileSet", "@id": "assets", "name": "precondition-assets", "description": "Pre-staged synthetic input files referenced by tasks (no personal data).", "encodingFormat": "application/octet-stream", "includes": "assets/*" } ], "recordSet": [ { "@type": "cr:RecordSet", "@id": "tasks", "name": "tasks", "description": "The 1448 evaluation tasks.", "key": { "@id": "tasks/uid" }, "field": [ { "@type": "cr:Field", "@id": "tasks/uid", "name": "uid", "description": "Globally unique task id (/).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "uid" } } }, { "@type": "cr:Field", "@id": "tasks/track", "name": "track", "description": "Source track: skill / mobile / mobile_agent / general.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "track" } } }, { "@type": "cr:Field", "@id": "tasks/category", "name": "category", "description": "Top-level category (anonymized).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "category" } } }, { "@type": "cr:Field", "@id": "tasks/subcategory", "name": "subcategory", "description": "Full category path (anonymized).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "subcategory" } } }, { "@type": "cr:Field", "@id": "tasks/id", "name": "id", "description": "Original task id within its track.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "id" } } }, { "@type": "cr:Field", "@id": "tasks/question", "name": "question", "description": "Natural-language task instruction for the agent.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "question" } } }, { "@type": "cr:Field", "@id": "tasks/reference_answer", "name": "reference_answer", "description": "Short reference / expected-outcome note.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "reference_answer" } } }, { "@type": "cr:Field", "@id": "tasks/checkpoints", "name": "checkpoints", "description": "Human-defined verification conditions (repeated).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "checkpoints" } }, "repeated": true }, { "@type": "cr:Field", "@id": "tasks/priority", "name": "priority", "description": "Priority P0-P3.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "priority" } } }, { "@type": "cr:Field", "@id": "tasks/difficulty", "name": "difficulty", "description": "easy / medium / hard.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "difficulty" } } }, { "@type": "cr:Field", "@id": "tasks/environment", "name": "environment", "description": "desktop_linux / mobile_android / empty.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "environment" } } }, { "@type": "cr:Field", "@id": "tasks/execution_mode", "name": "execution_mode", "description": "Execution mode, e.g. one_shot.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "execution_mode" } } }, { "@type": "cr:Field", "@id": "tasks/timeout", "name": "timeout", "description": "Suggested wall-clock budget in seconds.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "timeout" } } }, { "@type": "cr:Field", "@id": "tasks/step_budget", "name": "step_budget", "description": "Suggested maximum agent steps.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "step_budget" } } }, { "@type": "cr:Field", "@id": "tasks/token_budget", "name": "token_budget", "description": "Suggested token budget.", "dataType": "sc:Integer", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "token_budget" } } }, { "@type": "cr:Field", "@id": "tasks/precondition_files", "name": "precondition_files", "description": "JSON-encoded list of pre-staged input files under assets/.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "precondition_files" } } }, { "@type": "cr:Field", "@id": "tasks/tags", "name": "tags", "description": "Optional tags (repeated).", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "tags" } }, "repeated": true }, { "@type": "cr:Field", "@id": "tasks/extra_json", "name": "extra_json", "description": "JSON-encoded additional track-specific fields.", "dataType": "sc:Text", "source": { "fileObject": { "@id": "jsonl-file" }, "extract": { "column": "extra_json" } } } ] } ] }