{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-07-004",
    "source_story_id": "tmp-story-harbor-agent-evaluation-index",
    "edition_id": "mp-2026-09-07-morning-0060",
    "edition_url": "https://themachinepress.com/edition/2026-09-07",
    "position": 4,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "Eighty-Two Hard Tasks Cut Through Fifty-Four Agent Benchmarks",
    "slug": "eighty-two-hard-tasks-cut-through-fifty-four-agent-benchmarks",
    "dek": "Harbor Adapters ports more than 80 agent benchmarks into one infrastructure, while Harbor-Index distills 29 of them into a smaller audited set.",
    "summary": "Harbor Adapters ports more than 80 agent benchmarks into one infrastructure, while Harbor-Index distills 29 of them into a smaller audited set.",
    "body_text": "The project validates benchmark adapters through code review and parity experiments, then evaluates eight models across 54 benchmarks using a shared agent plus native harnesses. Harbor-Index selects 82 difficult, diverse tasks spanning 29 benchmarks after difficulty filtering and human-and-AI audit. No evaluated model-harness configuration exceeded 30 percent pass rate; the strongest reported result was 28.0 percent. The index lowers evaluation cost, but its conclusions still depend on the selected tasks, adapters and harnesses.",
    "why_it_matters": "Harbor Adapters ports more than 80 agent benchmarks into one infrastructure, while Harbor-Index distills 29 of them into a smaller audited set.",
    "limitations": [],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-07-004/eighty-two-hard-tasks-cut-through-fifty-four-agent-benchmarks",
    "json_url": "https://themachinepress.com/story/mp-2026-09-07-004.json",
    "first_published_at": "2026-09-07T09:00:00.000-04:00",
    "modified_at": "2026-09-07T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-07-004-001",
        "text": "Harbor Adapters ports more than 80 agent benchmarks into one infrastructure, while Harbor-Index distills 29 of them into a smaller audited set.",
        "source_ids": [
          "source-2026-09-07-004"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-09-07-004"
    ],
    "tags": [
      "agent benchmarks",
      "evaluation infrastructure",
      "Harbor-Index"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-07-004",
      "title": "arXiv preprint 2609.04298",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.04298",
      "canonical_url": "https://arxiv.org/abs/2609.04298",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-03T12:26:20.000-04:00",
      "accessed_at": "2026-09-07T08:20:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-07-004-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Eighty-Two Hard Tasks Cut Through Fifty-Four Agent Benchmarks",
    "publisher": "The Machine Press",
    "published_at": "2026-09-07T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-07-004/eighty-two-hard-tasks-cut-through-fifty-four-agent-benchmarks"
  }
}