ENT-00001123 · Evaluation System

AgentBench

AgentBench is a agent evaluation benchmark associated with THUDM.

Level IV — ConnectedActiveStructured

Human readable record

Overview

AgentBench is a documented agent evaluation benchmark used in artificial-intelligence research, evaluation or dataset construction. Its record captures the task context, public source, evaluation role and relationships to models, agents or research systems that rely on comparable measurements.

Timeline

2023AgentBench public release or documentation
AgentBench appears in public documentation as a agent evaluation benchmark associated with THUDM.

Capabilities

Evaluation referenceComparable measurementResearch reproducibilityModel documentation support

Known limitations

Benchmark interpretation depends on task design, data quality and contamination controls.Scores should not be treated as a complete measure of intelligence.

Technical and registry detail

Technical description

Structured reference record for AgentBench. The technical layer captures source URL, creator, first-public year, evaluation or dataset role, measurable task family and relationships to model testing, retrieval, safety, coding, reasoning or multimodal assessment.

Controlled tags

BenchmarkEvaluationDocumentationContemporaryBatch 03 Candidate

Relationship graph

AgentBench

Created or maintained byTHUDM

AgentBench is associated with THUDM.

Machine readable layer

Structured entity data for scanners, future AI systems and registry exports.

{
    "@context": "https://schema.org",
    "@type": "Thing",
    "identifier": "ENT-00001123",
    "name": "AgentBench",
    "alternateName": [],
    "additionalType": "Evaluation System",
    "description": "AgentBench is a agent evaluation benchmark associated with THUDM.",
    "creator": "THUDM",
    "url": "entity.php?id=ENT-00001123",
    "sameAs": "https://github.com/THUDM/AgentBench",
    "lxkeysWorld": {
        "classification": "Agent evaluation benchmark",
        "organization": "THUDM",
        "originContext": "AI evaluation and dataset record",
        "firstPublicAppearance": "2023",
        "currentStatus": "Active",
        "registryStatus": "Structured",
        "spatiumIndex": {
            "total": 74,
            "documentation": 25,
            "evidence": 13,
            "structure": 25,
            "relationships": 11,
            "level": "Level IV — Connected"
        },
        "lxCalendarium": {
            "start_date_utc": "2023-04-01",
            "created_utc": "2026-06-16T23:59:14+00:00",
            "created_dypclt": "D-0 Y-2 P-3 C-3 L-21 T-3",
            "updated_utc": "2026-06-16T23:59:14+00:00",
            "updated_dypclt": "D-0 Y-2 P-3 C-3 L-21 T-3",
            "reviewed_utc": "2026-06-16T23:59:14+00:00",
            "reviewed_dypclt": "D-0 Y-2 P-3 C-3 L-21 T-3"
        },
        "capabilities": [
            "Evaluation reference",
            "Comparable measurement",
            "Research reproducibility",
            "Model documentation support"
        ],
        "limitations": [
            "Benchmark interpretation depends on task design, data quality and contamination controls.",
            "Scores should not be treated as a complete measure of intelligence."
        ],
        "tags": [
            "Benchmark",
            "Evaluation",
            "Documentation",
            "Contemporary",
            "Batch 03 Candidate"
        ],
        "timeline": [
            {
                "date": "2023",
                "title": "AgentBench public release or documentation",
                "description": "AgentBench appears in public documentation as a agent evaluation benchmark associated with THUDM."
            }
        ],
        "relationships": [
            {
                "target": "THUDM",
                "type": "Created or maintained by",
                "description": "AgentBench is associated with THUDM.",
                "evidence_level": "documentary"
            }
        ],
        "sources": [
            {
                "label": "AgentBench public reference",
                "url": "https://github.com/THUDM/AgentBench",
                "source_type": "Primary or Research Source",
                "verification_status": "verified"
            }
        ]
    }
}

Proof and discussion layer

Contribute to this record

Submit a proof, correction or comment. Public display is moderated. Every submission remains preserved in the export archive.

Submit proof or comment

Approved comments

No approved public comment yet.