{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-15-008",
    "source_story_id": "tmp-story-elastic-transformer-budget",
    "edition_id": "mp-2026-10-15-morning-0097",
    "edition_url": "https://themachinepress.com/edition/2026-10-15",
    "position": 8,
    "story_type": "dispatch",
    "section": "infrastructure",
    "editorial_classification": "editorial",
    "headline": "One Transformer Slides Across Its Compute Budget",
    "slug": "one-transformer-slides-across-its-compute-budget",
    "dek": "A retraining-free ranking of individual neurons and attention heads lets one pretrained model trade capacity for speed across a continuum of sparsity targets.",
    "summary": "A retraining-free ranking of individual neurons and attention heads lets one pretrained model trade capacity for speed across a continuum of sparsity targets.",
    "body_text": "Model families normally force a fixed size choice even when hardware and load change. This method jointly ranks individual MLP neurons and attention heads while optimizing across many soft-pruning targets. Across 27 encoders from 22 million to 8 billion parameters, the authors report smooth degradation through 60 percent sparsity. At 50 percent sparsity, a DINOv3 encoder loses 427 million parameters with a 0.8-point linear-probe accuracy drop; another model reaches 1.80 times throughput at 60 percent sparsity. Those figures are workload-specific, but the output is one adjustable model rather than a shelf of separately trained sizes.",
    "why_it_matters": "A retraining-free ranking of individual neurons and attention heads lets one pretrained model trade capacity for speed across a continuum of sparsity targets.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-15-008/one-transformer-slides-across-its-compute-budget",
    "json_url": "https://themachinepress.com/story/mp-2026-10-15-008.json",
    "first_published_at": "2026-10-15T09:00:00.000-04:00",
    "modified_at": "2026-10-15T09:00:00.000-04:00",
    "content_status": "updated",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-15-008-001",
        "text": "A retraining-free ranking of individual neurons and attention heads lets one pretrained model trade capacity for speed across a continuum of sparsity targets.",
        "source_ids": [
          "source-2026-10-15-008"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-15-008"
    ],
    "tags": [
      "transformers",
      "pruning",
      "inference efficiency"
    ],
    "image_url": "https://themachinepress.com/issues/2026-10-15/elastic-transformer-compute.webp",
    "corrections": [
      {
        "correction_id": "correction-2026-10-15-001",
        "revision": 2,
        "summary": "Publication note: This October 15 edition was prepared before its scheduled 9:00 AM America/Detroit reveal, but publication was delayed by a production build failure. The original reporting and image-generation timestamps are preserved.",
        "published_at": "2026-10-15T09:00:00.000-04:00",
        "affected_story_ids": [],
        "affected_claim_ids": []
      }
    ]
  },
  "sources": [
    {
      "source_id": "source-2026-10-15-008",
      "title": "arXiv preprint 2610.20588, original v1",
      "publisher": "arXiv / submitting research authors",
      "url": "https://arxiv.org/abs/2610.20588v1",
      "canonical_url": "https://arxiv.org/abs/2610.20588v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-14T13:58:08.000-04:00",
      "accessed_at": "2026-10-15T08:28:29.903-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-15-008-001"
      ]
    }
  ],
  "corrections": [
    {
      "correction_id": "correction-2026-10-15-001",
      "revision": 2,
      "summary": "Publication note: This October 15 edition was prepared before its scheduled 9:00 AM America/Detroit reveal, but publication was delayed by a production build failure. The original reporting and image-generation timestamps are preserved.",
      "published_at": "2026-10-15T09:00:00.000-04:00",
      "affected_story_ids": [],
      "affected_claim_ids": []
    }
  ],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "One Transformer Slides Across Its Compute Budget",
    "publisher": "The Machine Press",
    "published_at": "2026-10-15T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-15-008/one-transformer-slides-across-its-compute-budget"
  }
}