{
  "id": 4406,
  "url": "https://arxiv.org/abs/2605.18848v4",
  "title": "Exact Linear Attention",
  "summary": "This paper introduces Exact Linear Attention (ELA), a mechanism that achieves linear computational complexity for Transformer attention by exploiting the exact decomposition property of kernel functions, thereby eliminating approximation error. We identify and address two key limitations of prior linear attention -- gradient explosion and token attention dilution -- by imposing kernel constraints that ensure non-negativity, discriminability, and geometric interpretability. Several kernel functio",
  "authors": "Weinuo Ou",
  "category": "research",
  "topics": "safety-alignment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-13T08:06:48.000Z",
  "fetched_at": "2026-07-14T16:30:59.236Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4406",
  "original_url": "https://arxiv.org/abs/2605.18848v4",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}