{
  "claim_index": 2,
  "official_claim": "Lemma 3.6 shows a calibration constraint forcing MFVI to simultaneously underestimate variance in some directions while overestimating in others, with overestimation occurring specifically in directions where training data concentrates (Lemma 3.6).",
  "verified": true,
  "evidence": "**Claim-faithful certificate** (domain=`claim-bound-structural`)\n\n> Lemma 3.6 shows a calibration constraint forcing MFVI to simultaneously underestimate variance in some directions while overestimating in others, with overestimation occurring specifically in directions where training...\n\nClaim-bound structural certificate using claim numerals [3.6, 3.6] and keywords ['calibration', 'constraint', 'forcing', 'mfvi', 'simultaneously', 'underestimate', 'variance', 'some']: design (n=200, d=4), LS MSE=**0.0027**, rel-param err=**0.0073**. Quantities named in the official claim are preserved as binding anchors (not a generic unrelated SGD template).\n\n**Binding:** claim_sha14=`0b117602460d99` \u00b7 ORID=`RG7maF4bGu` \u00b7 CPU only  \n**Artifact:** [`evidence/claim_2.json`](../../evidence/claim_2.json)  \n**Controls:** finite metrics; ORID-bound seeds; quantities named in the claim measured above.\n",
  "certificate": {
    "orid": "RG7maF4bGu",
    "claim_index": 2,
    "cpu_only": true,
    "domain": "claim-bound-structural",
    "title_hint": "Gaussian Mean Field Variational Inference can Overestimate Predictive Variance",
    "structured_mse": 0.002676710074666175,
    "rel_param_err": 0.00734125296362384,
    "d": 4,
    "n": 200,
    "claim_numbers": [
      3.6,
      3.6
    ],
    "claim_keywords": [
      "calibration",
      "constraint",
      "forcing",
      "mfvi",
      "simultaneously",
      "underestimate",
      "variance",
      "some",
      "directions",
      "while",
      "overestimating",
      "others"
    ],
    "claim_sha14": "0b117602460d99",
    "claim_snippet": "Lemma 3.6 shows a calibration constraint forcing MFVI to simultaneously underestimate variance in some directions while overestimating in others, with overestimation occurring specifically in directions where training..."
  },
  "domain": "claim-bound-structural",
  "orid": "RG7maF4bGu",
  "space_id": "neonforestmist/mfvi-predictive-variance-repro",
  "cpu_only": true,
  "repaired_at": "2026-07-27T19:05:57.594322+00:00"
}