{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-22-014",
    "source_story_id": "tmp-story-agent-behavior-maps",
    "edition_id": "mp-2026-10-22-morning-0104",
    "edition_url": "https://themachinepress.com/edition/2026-10-22",
    "position": 14,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "Agents Can Be Mapped Without Running New Rollouts",
    "slug": "agents-can-be-mapped-without-running-new-rollouts",
    "dek": "Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do.",
    "summary": "Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do.",
    "body_text": "Agent Behavior Maps compare action log-likelihoods on the same reference trajectories instead of launching each target agent in a fresh environment. Across eight cross-scored models, distances followed measured trajectory divergence in rank order. Across 100 language models and four environments, ALFWorld-derived coordinates predicted performance and exposed behavioral similarities that persisted after controlling for success rate and model size. The guarantees depend on explicit assumptions, and the map is not a substitute for deployment testing.",
    "why_it_matters": "Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do.",
    "limitations": [
      "The guarantees depend on explicit assumptions, and the map is not a substitute for deployment testing."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-22-014/agents-can-be-mapped-without-running-new-rollouts",
    "json_url": "https://themachinepress.com/story/mp-2026-10-22-014.json",
    "first_published_at": "2026-10-22T09:00:00.000-04:00",
    "modified_at": "2026-10-22T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-22-014-001",
        "text": "Centered action likelihoods on shared trajectories locate model-and-harness combinations by what they would do.",
        "source_ids": [
          "source-2026-10-22-014"
        ],
        "qualification": "The guarantees depend on explicit assumptions, and the map is not a substitute for deployment testing."
      }
    ],
    "source_ids": [
      "source-2026-10-22-014"
    ],
    "tags": [
      "agents",
      "evaluation",
      "behavior mapping"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-22-014",
      "title": "arXiv preprint 2610.30014, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.30014v1",
      "canonical_url": "https://arxiv.org/abs/2610.30014v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-21T12:44:02.000-04:00",
      "accessed_at": "2026-10-22T08:26:13.404-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-22-014-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Agents Can Be Mapped Without Running New Rollouts",
    "publisher": "The Machine Press",
    "published_at": "2026-10-22T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-22-014/agents-can-be-mapped-without-running-new-rollouts"
  }
}