{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "Profiling shows attention computation is a bottleneck in your self-hosted service.",
    "change": "Check the supported GPU, tensor shapes, and precision, then replace only the relevant kernel.",
    "baseline": "The existing kernel inside the same complete serving pipeline.",
    "steps": [
      "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.",
      "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.",
      "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.",
      "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test."
    ],
    "measure": "Numerical differences, end-to-end latency, throughput, and memory at realistic shapes.",
    "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation."
  },
  "builder_review": null,
  "arxiv_id": "2610.00564",
  "title": "Attention Kernels for Learning Maps Between Heavy-Tailed Measures",
  "summary": "The method replaces the exponential weighting in softmax attention with slower-growing kernel functions so that attention integrals remain finite when learning maps between probability measures with polynomial tails. The cost is minimal in memory and latency, but it requires a custom attention implementation rather than off-the-shelf fused softmax kernels, and it matters only for post-norm transformers trained on heavy-tailed ensembles. It can fail to help on light-tailed data, where all kernels perform equivalently, and even data transformation (symlog) does not rescue softmax on the sheared swap task, so kernel choice rather than preprocessing must carry the fix.",
  "rationale": "The gain is specific to operator learning on heavy-tailed measures and is validated only on two constructed benchmarks, so a practitioner should validate it on their own heavy-tailed workload before adopting it as a default.",
  "technique": "Slower-growing attention kernels for heavy-tailed operator learning",
  "family": "kernels_e_atencao",
  "recommendation": "testar",
  "published": "2026-09-30",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.00564",
  "independent_implementations": 0,
  "total_implementations": 0,
  "stars": 0,
  "citations": 0,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "The method replaces the exponential weighting in softmax attention with slower-growing kernel functions so that attention integrals remain finite when learning maps between probability measures with polynomial tails. The cost is minimal in memory and latency, but it requires a custom attention implementation rather than off-the-shelf fused softmax kernels, and it matters only for post-norm transformers trained on heavy-tailed ensembles. It can fail to help on light-tailed data, where all kernels perform equivalently, and even data transformation (symlog) does not rescue softmax on the sheared swap task, so kernel choice rather than preprocessing must carry the fix.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.00564",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "Slower-growing attention kernels avoid ensemble collapse on both heavy-tailed benchmarks where softmax collapses; symlog preprocessing rescues softmax only on one of two tasks; all kernels perform similarly on the Gaussian control.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.00564",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "exposure_map": [
    {
      "dimension": "quality",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compute",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "latency",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "operations",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "compatibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "security",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "data_and_training",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "dimension": "reproducibility",
      "basis": "not_evaluated",
      "finding": "",
      "interpretation": "",
      "limitation": "",
      "next_check": "",
      "source_url": "",
      "source_page": null,
      "source_excerpt": ""
    }
  ],
  "risks": [],
  "minimum_test": [],
  "open_questions": [],
  "independent_tests": [],
  "lab_reproductions": [],
  "report_available": false,
  "validation_tier": "unknown",
  "evidence_tier": "unknown",
  "infrastructure_basis": "unknown",
  "equations": [],
  "equations_status": "not_fetched",
  "core_kind": "",
  "equations_fetched_at": "",
  "equations_source_url": "",
  "equations_source_sha256": "",
  "exposures_reviewed_at": "",
  "exposures_source_sha256": ""
}