{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "Long customer conversations exhaust memory or limit concurrent requests.",
    "change": "Change only the attention cache policy, after checking model and runtime support.",
    "baseline": "The same model with its ordinary attention cache.",
    "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": "Peak cache memory, long-context answer quality, first-token latency, and requests completed.",
    "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": "2609.38121",
  "title": "WUSH-KV: KV Cache Quantization with Data-Adaptive Transforms",
  "summary": "WUSH-KV replaces generic or rotation-based transforms (e.g., OSCAR/Hadamard-style) in low-bit KV cache quantization with calibration-derived, second-order-statistics transforms for keys and values, folded into weights or applied after RoPE. It costs a calibration pass, minor weight modification for the value transform, and integration effort, since the reference implementation targets SGLang with OSCAR-style clipped affine quantization. It can fail if the calibration data is unrepresentative of production traffic, if the serving stack is not SGLang-compatible, or if perplexity gains do not translate to the reader's specific downstream tasks at their chosen bit-width.",
  "rationale": "The method addresses a real constraint (KV cache memory at long context) with modest engineering cost and no retraining, but its gains are validated only within a specific quantizer and serving stack, so the reader should validate on their own models and workloads before production adoption.",
  "technique": "Data-adaptive transforms for 2-bit KV cache quantization",
  "family": "cache_kv",
  "recommendation": "testar",
  "published": "2026-09-29",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2609.38121",
  "independent_implementations": 0,
  "total_implementations": 1,
  "stars": 1,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "WUSH-KV replaces generic or rotation-based transforms (e.g., OSCAR/Hadamard-style) in low-bit KV cache quantization with calibration-derived, second-order-statistics transforms for keys and values, folded into weights or applied after RoPE. It costs a calibration pass, minor weight modification for the value transform, and integration effort, since the reference implementation targets SGLang with OSCAR-style clipped affine quantization. It can fail if the calibration data is unrepresentative of production traffic, if the serving stack is not SGLang-compatible, or if perplexity gains do not translate to the reader's specific downstream tasks at their chosen bit-width.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.38121",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "At 2-bit, WUSH-KV performs comparably to or outperforms the OSCAR transform across all evaluated models and downstream tasks, with the lowest end-to-end perplexity among tested transforms; no multiplicative factor is reported.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2609.38121",
      "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": ""
}