{
  "annotation_id": "5dee277f-914e-4a8d-af70-0b72b155e65d",
  "date_created": "2026-02-17T05:53:12.439000Z",
  "date_modified": "2026-02-17T05:53:12.439000Z",
  "file_hash": "2519a4645593e4dc720786ab3d119d764ab4381cdf6c176a2ebad9c86e6fec05",
  "private": false,
  "record": {
    "abstract": "We propose score-based VAMP (SC-VAMP), a variant of vector approximate message passing (VAMP) in which the Onsager correction is expressed and computed via conditional Fisher information, thereby enabling a Jacobian-free implementation. Using learned score functions, SC-VAMP constructs nonlinear MMSE estimators through Tweedie\u0027s formula and derives the corresponding Onsager terms from the score-norm statistics, avoiding the need for analytical derivatives of the prior or likelihood. When combined with random orthogonal/unitary mixing to mitigate non-ideal, structured or correlated sensing settings, the proposed framework extends VAMP to complex black-box inference problems where explicit modeling is intractable. Finally, by leveraging the entropic CLT, we provide an information-theoretic perspective on the Gaussian approximation underlying SE, offering insight into the decoupling principle beyond idealized i.i.d. settings, including nonlinear regimes.",
    "arxiv_id": "2601.07095",
    "authors": [
      "Tadashi Wadayama",
      "Takumi Takahashi"
    ],
    "categories": [
      "cs.IT",
      "math.IT"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "title": "Score-Based VAMP with Fisher-Information-Based Onsager Correction",
    "url": "https://arxiv.org/abs/2601.07095",
    "version": "v1"
  },
  "schema_id": "dorsal/arxiv",
  "source": {
    "execution_id": "7c72ce07-7c56-4e9c-8d95-b880d673ac2c",
    "id": "arXiv Dataset",
    "type": "Model",
    "variant": "snapshot-2026-01-17",
    "version": "0.1.0"
  },
  "user_id": 1000002
}