{ "schema_version": 1, "builder_plan": { "scenario": "You already need task-specific training and want to reduce its memory or compute cost.", "change": "Change only the training technique on a small controlled run after checking prerequisites.", "baseline": "Your current training recipe with the same model, data split, and evaluation.", "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": "Held-out quality, peak memory, training time, and full compute cost to the target quality.", "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": "2609.15152", "title": "PACE: Progressive Angular-to-Norm Contrastive Embedding", "summary": "PACE replaces single-stage cosine-based contrastive fine-tuning of multimodal embedding models with a two-stage schedule: LoRA with a cosine objective to fix angular geometry, then full-parameter fine-tuning with a dot-product objective plus a confidence-adaptive focal loss so embedding norms carry semantic signal. The cost is a more complex two-phase training pipeline and a Stage II full-parameter pass that, depending on backbone size, may exceed a single 24 GB GPU without gradient checkpointing or offloading. It can fail if the angular space established in Stage I is poorly formed, since Stage II amplifies that geometry, and the added hyperparameters (stage switch point, focal loss weighting) require per-task tuning.", "rationale": "The method is a fine-tuning recipe the reader could run on one GPU for smaller backbones, but the unquantified gains and the full-parameter Stage II require workload-specific validation before production adoption.", "technique": "Progressive angular-to-norm contrastive embedding training", "family": "treino_eficiente", "recommendation": "testar", "published": "2026-09-14", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.15152", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "PACE replaces single-stage cosine-based contrastive fine-tuning of multimodal embedding models with a two-stage schedule: LoRA with a cosine objective to fix angular geometry, then full-parameter fine-tuning with a dot-product objective plus a confidence-adaptive focal loss so embedding norms carry semantic signal. The cost is a more complex two-phase training pipeline and a Stage II full-parameter pass that, depending on backbone size, may exceed a single 24 GB GPU without gradient checkpointing or offloading. It can fail if the angular space established in Stage I is poorly formed, since Stage II amplifies that geometry, and the added hyperparameters (stage switch point, focal loss weighting) require per-task tuning.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.15152", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "The abstract reports consistent improvements over cosine-only contrastive training across multiple backbone scales and multimodal embedding tasks, but gives no quantified figure; gains must be verified against the full paper's tables.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.15152", "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_formula", "core_kind": "algorithm", "equations_fetched_at": "2026-10-05", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }