Pith Number

pith:TNSPLWAX

pith:2026:TNSPLWAXCQM7WHFNBI3NUMIQ7Q

not attested not anchored not stored refs pending

Diffusion Model's Generalization Can Be Characterized by Inductive Biases toward a Data-Dependent Ridge Manifold

Molei Tao, Ye He, Yitong Qiu

Diffusion model samples evolve by reaching a data ridge, then aligning via normal error and sliding via tangential error.

arxiv:2602.06021 v2 · 2026-02-05 · stat.ML · cs.LG · cs.NA · math.NA · math.PR

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{TNSPLWAXCQM7WHFNBI3NUMIQ7Q}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Our main result shows that generated samples evolve by a reach-align-slide mechanism: they first enter a neighborhood of the ridge, then their distance to the ridge is controlled by the normal component of training error, and finally their motion along the ridge is controlled by the tangential component.

C2weakest assumption

That the time-dependent family of log-density ridge manifolds constructed from the smoothed empirical distribution accurately captures the relevant geometry for characterizing reverse-time inference without introducing artifacts that alter the predicted reach-align-slide behavior.

C3one line summary

Diffusion model generated samples follow a reach-align-slide path on data-dependent ridge manifolds, with normal and tangential training error components controlling distance to and motion along the ridge.

Formal links

2 machine-checked theorem links

Cited by

2 papers in Pith

The finite expression method for turbulent dynamics with high-order moment recovery

Christoffel-DPS: Optimal sensor placement in diffusion posterior sampling for arbitrary distributions

Receipt and verification

First computed	2026-05-18T02:44:31.384415Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

9b64f5d8171419fb1cad0a36da3110fc0c6fca3823593454606fd49cb58962dd

Aliases

arxiv: 2602.06021 · arxiv_version: 2602.06021v2 · doi: 10.48550/arxiv.2602.06021 · pith_short_12: TNSPLWAXCQM7 · pith_short_16: TNSPLWAXCQM7WHFN · pith_short_8: TNSPLWAX

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/TNSPLWAXCQM7WHFNBI3NUMIQ7Q \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 9b64f5d8171419fb1cad0a36da3110fc0c6fca3823593454606fd49cb58962dd

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "2868ced2af0d679389848356f189234bd426ab88e86462d50c297b9b6282b978",
    "cross_cats_sorted": [
      "cs.LG",
      "cs.NA",
      "math.NA",
      "math.PR"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "stat.ML",
    "submitted_at": "2026-02-05T18:55:03Z",
    "title_canon_sha256": "14a1717bd039ead751343f9d4b08a277b75d84eb848ca6ad71990e0ea6c3e245"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2602.06021",
    "kind": "arxiv",
    "version": 2
  }
}