{
  "schema_version": 1,
  "builder_plan": {
    "scenario": "A narrow, repeated product task may not need your largest model.",
    "change": "Check whether a released student model fits the task before generating training data.",
    "baseline": "Your current model and an off-the-shelf small model on the same held-out tasks.",
    "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": "Student quality, serving cost, teacher calls, training cost, and failures on unusual inputs.",
    "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": "2610.06671",
  "title": "Learning What to Imitate: Entropy-Aware Distribution Mixing",
  "summary": "Replaces fixed teacher-supervised token-level imitation (standard SFT on teacher traces or fixed-target forward-KL distillation) with per-token interpolation between student and teacher distributions, gated by the student's predictive entropy. Costs additional forward passes from both distributions at trace-generation or training time (speculative decoding for offline traces), plus implementation complexity of schedule selection; inference cost is unchanged. The optimal entropy schedule is source-dependent (concave for offline traces, linear or convex for on-policy), so adopting the wrong schedule for one's data regime may yield little benefit or residual degradation in generalisation.",
  "rationale": "The method is implementable on a single 24 GB GPU using open-weight student and teacher models, but its benefit depends on choosing an entropy schedule matched to the training data source, which requires workload-specific validation before production adoption.",
  "technique": "Entropy-gated student-teacher distribution mixing for distillation",
  "family": "destilacao",
  "recommendation": "testar",
  "published": "2026-10-05",
  "as_of": "2026-10-18",
  "editorial_status": "indexed",
  "source_url": "https://arxiv.org/abs/2610.06671",
  "independent_implementations": 0,
  "total_implementations": 1,
  "stars": 1,
  "citations": null,
  "claims": [
    {
      "claim_id": "claim-01",
      "statement": "Replaces fixed teacher-supervised token-level imitation (standard SFT on teacher traces or fixed-target forward-KL distillation) with per-token interpolation between student and teacher distributions, gated by the student's predictive entropy. Costs additional forward passes from both distributions at trace-generation or training time (speculative decoding for offline traces), plus implementation complexity of schedule selection; inference cost is unchanged. The optimal entropy schedule is source-dependent (concave for offline traces, linear or convex for on-policy), so adopting the wrong schedule for one's data regime may yield little benefit or residual degradation in generalisation.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.06671",
      "source_page": null,
      "source_excerpt": ""
    },
    {
      "claim_id": "claim-02",
      "statement": "Improves in-distribution and out-of-distribution math reasoning and better preserves general capabilities than fixed-teacher supervision; no numerical magnitude is reported in the abstract.",
      "basis": "inferred",
      "result": "",
      "baseline": "",
      "conditions": "",
      "source_url": "https://arxiv.org/abs/2610.06671",
      "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": ""
}