{
  "schema_version": 1,
  "title": "Reproduction: AsyncSpade: Efficient Test-Time Scaling with Asynchronous Sparse Decoding",
  "emoji": "🎯",
  "space_id": "debajyotidasgupta/repro-asyncspade-efficient-test-time-scaling-with-asynchronous-sparse-decoding",
  "paper": {
    "arxiv_id": "2510.07486"
  },
  "tags": [
    "icml2026-repro",
    "paper-avVTEzRgba"
  ],
  "updated_at": "2026-07-19T15:59:52+00:00",
  "root": {
    "slug": "index",
    "title": "Reproduction: AsyncSpade: Efficient Test-Time Scaling with Asynchronous Sparse Decoding",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "executive-summary",
        "title": "Executive summary",
        "file": "pages/executive-summary/page.md",
        "children": []
      },
      {
        "slug": "claim-1-asyncspade-achieves-over-20-reduction-in-time-per-output-token-tpot-compared-to-the-sota-sparse-attention-baseline-quest-evaluated-on-qwen3-8b-and-qwen3-32b-section-5",
        "title": "Claim 1: AsyncSpade achieves over 20% reduction in time-per-output-token (TPOT) compared to the SoTA sparse-attention baseline Quest, evaluated on Qwen3-8B and Qwen3-32B (Section 5).",
        "file": "pages/claim-1-asyncspade-achieves-over-20-reduction-in-time-per-output-token-tpot-compared-to-the-sota-sparse-attention-baseline-quest-evaluated-on-qwen3-8b-and-qwen3-32b-section-5/page.md",
        "children": []
      },
      {
        "slug": "claim-2-asyncspade-achieves-at-least-50-tpot-reduction-compared-to-full-attention-while-matching-or-surpassing-accuracy-on-aime-24-aime-25-gpqa-diamond-and-math-500-figure-9",
        "title": "Claim 2: AsyncSpade achieves at least 50% TPOT reduction compared to full attention while matching or surpassing accuracy on AIME-24, AIME-25, GPQA-Diamond, and MATH-500 (Figure 9).",
        "file": "pages/claim-2-asyncspade-achieves-at-least-50-tpot-reduction-compared-to-full-attention-while-matching-or-surpassing-accuracy-on-aime-24-aime-25-gpqa-diamond-and-math-500-figure-9/page.md",
        "children": []
      },
      {
        "slug": "claim-3-asyncspade-fully-overlaps-kv-cache-management-operations-with-the-inference-pipeline-within-a-defined-workload-range-achieving-the-theoretical-optimal-tpot-section-4-figures-4-5",
        "title": "Claim 3: AsyncSpade fully overlaps KV-cache management operations with the inference pipeline within a defined workload range, achieving the theoretical optimal TPOT (Section 4, Figures 4-5).",
        "file": "pages/claim-3-asyncspade-fully-overlaps-kv-cache-management-operations-with-the-inference-pipeline-within-a-defined-workload-range-achieving-the-theoretical-optimal-tpot-section-4-figures-4-5/page.md",
        "children": []
      },
      {
        "slug": "claim-4-cache-management-latency-and-required-bandwidth-are-measured-at-3-92-14-43ms-across-configurations-on-a100-and-h100-8-gpu-nodes-compared-to-inference-latencies-of-5-47-14-26ms-table-1-table-2",
        "title": "Claim 4: Cache management latency and required bandwidth are measured at 3.92-14.43ms across configurations on A100 and H100 8-GPU nodes, compared to inference latencies of 5.47-14.26ms (Table 1, Table 2).",
        "file": "pages/claim-4-cache-management-latency-and-required-bandwidth-are-measured-at-3-92-14-43ms-across-configurations-on-a100-and-h100-8-gpu-nodes-compared-to-inference-latencies-of-5-47-14-26ms-table-1-table-2/page.md",
        "children": []
      },
      {
        "slug": "claim-5-the-methods-query-prediction-component-exploits-observed-temporal-locality-and-linear-correlation-in-attention-patterns-across-decoding-steps-to-select-sparse-kv-subsets-asynchronously-section-3-figures-4-5",
        "title": "Claim 5: The methods query-prediction component exploits observed temporal locality and linear correlation in attention patterns across decoding steps to select sparse KV subsets asynchronously (Section 3, Figures 4-5).",
        "file": "pages/claim-5-the-methods-query-prediction-component-exploits-observed-temporal-locality-and-linear-correlation-in-attention-patterns-across-decoding-steps-to-select-sparse-kv-subsets-asynchronously-section-3-figures-4-5/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "Conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      }
    ]
  },
  "agent_view_tokens": 6984,
  "revision": "1784476792109701000"
}