{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-24-017",
    "source_story_id": "tmp-brief-action-learning-simulations",
    "edition_id": "mp-2026-10-24-morning-0106",
    "edition_url": "https://themachinepress.com/edition/2026-10-24",
    "position": 19,
    "story_type": "ticker",
    "section": "research",
    "editorial_classification": "editorial",
    "headline": "A Learned Action Scores Whole Scientific Trajectories",
    "slug": "a-learned-action-scores-whole-scientific-trajectories",
    "dek": "LEAP turns observed paths into one differentiable objective reusable for ranking, steering, constrained optimization and sampling.",
    "summary": "LEAP turns observed paths into one differentiable objective reusable for ranking, steering, constrained optimization and sampling.",
    "body_text": "Rather than learn only local transitions, the method compares observed trajectories with a tractable reference process and learns a correction on path space. Tests span stochastic physical systems, orbital dynamics and molecular simulations. The evidence is computational and does not establish every system’s governing action.",
    "why_it_matters": "LEAP turns observed paths into one differentiable objective reusable for ranking, steering, constrained optimization and sampling.",
    "limitations": [
      "Rather than learn only local transitions, the method compares observed trajectories with a tractable reference process and learns a correction on path space.",
      "The evidence is computational and does not establish every system’s governing action."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-24-017/a-learned-action-scores-whole-scientific-trajectories",
    "json_url": "https://themachinepress.com/story/mp-2026-10-24-017.json",
    "first_published_at": "2026-10-24T09:00:00.000-04:00",
    "modified_at": "2026-10-24T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-24-017-001",
        "text": "LEAP turns observed paths into one differentiable objective reusable for ranking, steering, constrained optimization and sampling.",
        "source_ids": [
          "source-2026-10-24-019"
        ],
        "qualification": "Rather than learn only local transitions, the method compares observed trajectories with a tractable reference process and learns a correction on path space."
      }
    ],
    "source_ids": [
      "source-2026-10-24-019"
    ],
    "tags": [
      "scientific computing",
      "trajectories",
      "simulation"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-24-019",
      "title": "arXiv preprint 2610.31585, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.31585v1",
      "canonical_url": "https://arxiv.org/abs/2610.31585v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-22T13:22:08.000-04:00",
      "accessed_at": "2026-10-24T08:23:48.747-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-24-017-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "A Learned Action Scores Whole Scientific Trajectories",
    "publisher": "The Machine Press",
    "published_at": "2026-10-24T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-24-017/a-learned-action-scores-whole-scientific-trajectories"
  }
}