{ "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": "2610.29544", "title": "RISK-KV: Quality-Aware Selective KV Cache Reconstruction for Distributed LLM Inference", "summary": "RISK-KV replaces full KV cache retention/transfer and full token replay during request migration and interruption recovery, reconstructing missing layers' KV states via cross-layer ridge regression folded into the model's projection matrices. It costs a topology search with quality gating, per-layer-group predictor fitting, storage for regression parameters, and a measured NLL degradation that must stay under the gating threshold. It can fail if reconstruction error exceeds the quality gate on out-of-distribution inputs, if the 512-token-history speedup does not generalize to longer contexts, or if the reader's workload lacks migration/recovery events frequent enough to amortize the search and fitting overhead.", "rationale": "The gain is validated only for a narrow recovery scenario on three models, so it merits workload-specific testing rather than direct adoption in a single-GPU or API-based setup.", "technique": "Quality-aware selective KV cache reconstruction", "family": "cache_kv", "recommendation": "testar", "published": "2026-10-21", "as_of": "2026-10-22", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.29544", "independent_implementations": 1, "total_implementations": 2, "stars": 4, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "RISK-KV replaces full KV cache retention/transfer and full token replay during request migration and interruption recovery, reconstructing missing layers' KV states via cross-layer ridge regression folded into the model's projection matrices. It costs a topology search with quality gating, per-layer-group predictor fitting, storage for regression parameters, and a measured NLL degradation that must stay under the gating threshold. It can fail if reconstruction error exceeds the quality gate on out-of-distribution inputs, if the 512-token-history speedup does not generalize to longer contexts, or if the reader's workload lacks migration/recovery events frequent enough to amortize the search and fitting overhead.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.29544", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "1.51x to 3.31x speedup over full Token Replay in request recovery with a 512-token history, while omitting 18.2% to 33.3% of historical KV with perplexity changes below 0.5%.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.29544", "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": "" }