{ "schema_version": 1, "builder_plan": { "scenario": "A product assistant retrieves plausible documents but gives unsupported answers.", "change": "Change one retrieval or ranking step while keeping the document snapshot fixed.", "baseline": "Your current retrieval pipeline with identical questions and answer model.", "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": "Supported answers, retrieval misses, correct abstentions, index cost, and response latency.", "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": "2610.13277", "title": "Beyond the Final Embedding: Retrieving Memories from Token-Level Evidence", "summary": "The method replaces single-vector query-memory similarity with sparse token-level matching over the contextual token representations already produced by the embedding model, aggregated and combined with whole-memory similarity, optionally refined by a low-rank query adapter trained while the encoder and stored representations stay fixed. The training-free variant costs only token-level index storage and per-query matching latency, but storing token representations for all memories is substantially larger than one vector per memory, and the adapter variant adds a supervised training step with labeled data. It can fail when query-token matches fire on irrelevant memories (raising false positives), when token index size or scoring latency exceeds the memory retrieval budget, and when gains on the eight reported benchmarks do not transfer to the reader's memory distribution.", "rationale": "The training-free variant is cheap to prototype on constrained infrastructure since it reuses existing embedding-model outputs, but the storage and latency overhead of token-level indexes plus the need to validate retrieval gains on the reader's own memory workload make workload-specific testing necessary before adoption.", "technique": "Token-level sparse matching for memory retrieval", "family": "recuperacao_e_rag", "recommendation": "testar", "published": "2026-10-09", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.13277", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces single-vector query-memory similarity with sparse token-level matching over the contextual token representations already produced by the embedding model, aggregated and combined with whole-memory similarity, optionally refined by a low-rank query adapter trained while the encoder and stored representations stay fixed. The training-free variant costs only token-level index storage and per-query matching latency, but storing token representations for all memories is substantially larger than one vector per memory, and the adapter variant adds a supervised training step with labeled data. It can fail when query-token matches fire on irrelevant memories (raising false positives), when token index size or scoring latency exceeds the memory retrieval budget, and when gains on the eight reported benchmarks do not transfer to the reader's memory distribution.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.13277", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "With an optional trained low-rank query adapter on Qwen3-Embedding-4B, average Recall@10 improves by 13.3 points and NDCG@10 by 14.3 points across eight benchmarks; training-free gains are smaller and the paper does not quantify them as a factor. The paper also reports higher QA accuracy on long-term memory benchmarks without a single headline figure.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.13277", "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": "" }