{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-08-16-010",
    "source_story_id": "tmp-story-tree-coupled-abn-testing",
    "edition_id": "mp-2026-08-16-morning-0038",
    "edition_url": "https://themachinepress.com/edition/2026-08-16",
    "position": 10,
    "story_type": "dispatch",
    "section": "research",
    "editorial_classification": "editorial",
    "headline": "Many A/B Tests Shared the Same Reward",
    "slug": "many-a-b-tests-shared-the-same-reward",
    "dek": "A tree-coupled design preserved each policy's standalone trajectory law while reusing matched feedback across comparisons.",
    "summary": "A tree-coupled design preserved each policy's standalone trajectory law while reusing matched feedback across comparisons.",
    "body_text": "Directly comparing J adaptive policies for T rounds consumes JT reward-bearing interactions. The proposed exact coupling connects policy histories with a predictable tree, shares one reward within matched components and retains each policy's finite-horizon law. Its query count becomes T plus cumulative edge mismatches and can approach T rather than JT when policies converge. Experiments on reward models, language-model evaluation and adaptive search reported a better cost-precision frontier; practical gains depend on how closely the policies' actions can be coupled.",
    "why_it_matters": "A tree-coupled design preserved each policy's standalone trajectory law while reusing matched feedback across comparisons.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-16-010/many-a-b-tests-shared-the-same-reward",
    "json_url": "https://themachinepress.com/story/mp-2026-08-16-010.json",
    "first_published_at": "2026-08-16T09:00:00.000-04:00",
    "modified_at": "2026-08-16T15:43:36.912-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-08-16-010-001",
        "text": "A tree-coupled design preserved each policy's standalone trajectory law while reusing matched feedback across comparisons.",
        "source_ids": [
          "source-2026-08-16-010"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-08-16-010"
    ],
    "tags": [
      "experimentation",
      "bandits",
      "A/B testing",
      "feedback sharing"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-08-16-010",
      "title": "arXiv preprint 2608.12831",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.12831",
      "canonical_url": "https://arxiv.org/abs/2608.12831",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-13T01:10:43.000-04:00",
      "accessed_at": "2026-08-16T15:41:08.010-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-08-16-010-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Many A/B Tests Shared the Same Reward",
    "publisher": "The Machine Press",
    "published_at": "2026-08-16T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-16-010/many-a-b-tests-shared-the-same-reward"
  }
}