{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-08-10-009",
    "source_story_id": "tmp-story-coinrag-nugget-cache",
    "edition_id": "mp-2026-08-10-morning-0032",
    "edition_url": "https://themachinepress.com/edition/2026-08-10",
    "position": 9,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "RAG Kept the Nuggets and Dropped the Chunks",
    "slug": "rag-kept-the-nuggets-and-dropped-the-chunks",
    "dek": "CoinRAG reuses cached, query-relevant semantic units instead of repeatedly encoding full retrieved passages.",
    "summary": "CoinRAG reuses cached, query-relevant semantic units instead of repeatedly encoding full retrieved passages.",
    "body_text": "CoinRAG first retrieves broad chunks, then identifies smaller information nuggets and assembles their precomputed key-value representations with compact context. On LongBench multi-hop question-answering tasks, the authors report lower prefill costs and a 5.3 percent average relative F1 improvement under a standard fast-latency budget compared with their baselines. The result is an efficiency claim on selected benchmarks, not proof that fine-grained cache reuse preserves every source nuance.",
    "why_it_matters": "CoinRAG reuses cached, query-relevant semantic units instead of repeatedly encoding full retrieved passages.",
    "limitations": [
      "The result is an efficiency claim on selected benchmarks, not proof that fine-grained cache reuse preserves every source nuance."
    ],
    "importance": 7,
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-10-009/rag-kept-the-nuggets-and-dropped-the-chunks",
    "json_url": "https://themachinepress.com/story/mp-2026-08-10-009.json",
    "first_published_at": "2026-08-10T09:00:00.000-04:00",
    "modified_at": "2026-08-10T15:50:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-08-10-009-001",
        "text": "CoinRAG reuses cached, query-relevant semantic units instead of repeatedly encoding full retrieved passages.",
        "source_ids": [
          "source-2026-08-10-009"
        ],
        "qualification": "The result is an efficiency claim on selected benchmarks, not proof that fine-grained cache reuse preserves every source nuance."
      }
    ],
    "source_ids": [
      "source-2026-08-10-009"
    ],
    "tags": [
      "RAG",
      "KV cache",
      "long context",
      "retrieval"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-08-10-009",
      "title": "arXiv preprint 2608.07458",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.07458",
      "canonical_url": "https://arxiv.org/abs/2608.07458",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-07T13:00:00.000-04:00",
      "accessed_at": "2026-08-10T15:30:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-08-10-009-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "RAG Kept the Nuggets and Dropped the Chunks",
    "publisher": "The Machine Press",
    "published_at": "2026-08-10T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-10-009/rag-kept-the-nuggets-and-dropped-the-chunks"
  }
}