{ "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": "2610.00664", "title": "PhysicsMate: A Curriculum-Grounded Bengali Benchmark for Secondary Physics QA with Small-Model Adaptation", "summary": "The method replaces prompting a general-purpose base model (or paying for a large API model) for domain-specific curriculum QA with LoRA adapters on small open models, optionally quantized to an offline binary. Cost is modest: a unified LoRA recipe on 0.6B-4B models fits a single 24 GB GPU, plus the effort of building curriculum-grounded training data (here via a knowledge graph of 1760 nodes); quantization for deployment trades a small amount of quality for a local offline artifact. It can fail on loosely specified entity-level knowledge, which the paper shows benefits least from adaptation, and gains are domain-specific: the recipe requires curated, curriculum-aligned data and may not transfer to other subjects or languages without equivalent data work.", "rationale": "LoRA on sub-4B models is directly executable on the reader's single 24 GB GPU, shows large quantified accuracy gains for domain QA, and the quantized offline binary matches a constrained-infrastructure production path.", "technique": "LoRA fine-tuning of small models for curriculum QA", "family": "treino_eficiente", "recommendation": "adotar", "published": "2026-09-30", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.00664", "independent_implementations": 0, "total_implementations": 1, "stars": 121, "citations": 0, "claims": [ { "claim_id": "claim-01", "statement": "The method replaces prompting a general-purpose base model (or paying for a large API model) for domain-specific curriculum QA with LoRA adapters on small open models, optionally quantized to an offline binary. Cost is modest: a unified LoRA recipe on 0.6B-4B models fits a single 24 GB GPU, plus the effort of building curriculum-grounded training data (here via a knowledge graph of 1760 nodes); quantization for deployment trades a small amount of quality for a local offline artifact. It can fail on loosely specified entity-level knowledge, which the paper shows benefits least from adaptation, and gains are domain-specific: the recipe requires curated, curriculum-aligned data and may not transfer to other subjects or languages without equivalent data work.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.00664", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "LoRA adaptation raised closed-book accuracy by +5.5, +15.0, and +23.3 percentage points at 0.6B, 1.7B, and 4B parameters respectively on a Bengali physics QA benchmark of 1834 pairs.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.00664", "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": "" }