pith. sign in
Pith Number

pith:NTVX2QFG

pith:2024:NTVX2QFGGXF3YBDNID45X54RGW
not attested not anchored not stored refs resolved

Orthogonal Subspace Decomposition for Generalizable AI-Generated Image Detection

Baoyuan Wu, Chengchun Liu, Jiangming Wang, Ke-Yue Zhang, Li Yuan, Peng Jin, Shen Chen, Shouhong Ding, Taiping Yao, Zhiyuan Yan

Decomposing features via SVD into orthogonal parts lets detectors freeze general pre-trained knowledge and adapt only the rest to spot AI fakes without overfitting.

arxiv:2411.15633 v4 · 2024-11-23 · cs.CV

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{NTVX2QFGGXF3YBDNID45X54RGW}

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

By employing Singular Value Decomposition (SVD) to decompose the original feature space into two orthogonal subspaces and freezing the principal components while adapting only the remained components, we preserve the pre-trained knowledge while learning fake patterns, effectively minimizing overfitting and enhancing generalization.

C2weakest assumption

The principal components obtained from SVD on features of pre-trained vision foundation models capture general knowledge that remains useful and orthogonal to the specific patterns needed for detecting fakes, such that freezing them does not remove information critical for the detection task.

C3one line summary

Orthogonal subspace decomposition via SVD on vision foundation model features preserves high-rank pre-trained knowledge by freezing principal components and adapting residuals, reducing overfitting for better generalization in AI-generated image detection.

References

290 extracted · 207 resolved · 16 Pith anchors

[1] Wukong, 2022. 5. In https://xihe.mindspore.cn/modelzoo/wukong, 2022. 5 2022
[3] Brock, A. et al. Large scale gan training for high fidelity natural image synthesis. In ICLR, 2018 b 2018
[4] End-to-end reconstruction-classification learning for face forgery detection 2022
[5] What makes fake images detectable? understanding properties that generalize 2020
[6] Drct: Diffusion reconstruction contrastive training towards universal detection of diffusion generated images 2024

Formal links

2 machine-checked theorem links

Cited by

22 papers in Pith

Receipt and verification
First computed 2026-05-17T23:37:42.503015Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

6ceb7d40a635cbbc046d40f9dbf791359730559d773de73f6faa62963527dde1

Aliases

arxiv: 2411.15633 · arxiv_version: 2411.15633v4 · doi: 10.48550/arxiv.2411.15633 · pith_short_12: NTVX2QFGGXF3 · pith_short_16: NTVX2QFGGXF3YBDN · pith_short_8: NTVX2QFG
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/NTVX2QFGGXF3YBDNID45X54RGW \
  | 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: 6ceb7d40a635cbbc046d40f9dbf791359730559d773de73f6faa62963527dde1
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "439fd46a386a914a6a76f029ae6262bf358b60d175f9cd814cb6f494a0d828ed",
    "cross_cats_sorted": [],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CV",
    "submitted_at": "2024-11-23T19:10:32Z",
    "title_canon_sha256": "8562b9d00d4e88e6d8d323576b11263ac5229bdb152520a874df7728dfa341ae"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2411.15633",
    "kind": "arxiv",
    "version": 4
  }
}