{ "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": "2609.27262", "title": "Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs", "summary": "This replaces the common practice of maintaining separate specialist fine-tunes (or ad-hoc sequential updates) per support skill with a single multi-task fully fine-tuned checkpoint per model size. The cost is one full fine-tuning run over the combined task data and full-precision optimizer memory during training, which on a 24 GB GPU limits feasible sizes to small models or requires API-based fine-tuning for larger ones; sequential LoRA is the fallback when full fine-tuning is infeasible and better preserves earlier skills than sequential full fine-tuning. Failure modes include sharp off-task degradation if specialists are deployed without a reliable router, and catastrophic forgetting of earlier skills under sequential full fine-tuning when tasks are added over time.", "rationale": "The paper provides directly actionable, size-independent guidance—default to one multi-task fully fine-tuned model, use LoRA for sequential updates, and merge with the base model to restore robustness—which a small team can apply immediately within a single-GPU or API budget.", "technique": "Multi-task full fine-tuning as deployment default", "family": "treino_eficiente", "recommendation": "adotar", "published": "2026-09-23", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2609.27262", "independent_implementations": 0, "total_implementations": 0, "stars": 0, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "This replaces the common practice of maintaining separate specialist fine-tunes (or ad-hoc sequential updates) per support skill with a single multi-task fully fine-tuned checkpoint per model size. The cost is one full fine-tuning run over the combined task data and full-precision optimizer memory during training, which on a 24 GB GPU limits feasible sizes to small models or requires API-based fine-tuning for larger ones; sequential LoRA is the fallback when full fine-tuning is infeasible and better preserves earlier skills than sequential full fine-tuning. Failure modes include sharp off-task degradation if specialists are deployed without a reliable router, and catastrophic forgetting of earlier skills under sequential full fine-tuning when tasks are added over time.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.27262", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Across thirteen models (0.6B–32B) and eight customer-support datasets, multi-task full fine-tuning was the strongest performer at every model size tested; specialists degraded sharply off-task, and merging a specialist with its base model recovered off-task robustness with limited same-task loss for larger models. No multiplicative factor is reported in the abstract.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2609.27262", "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": "" }