Pith Number

pith:TKNCVVC7

pith:2026:TKNCVVC76J3KRK2NEPNL3GJYPY

not attested not anchored not stored refs resolved

To See is Not to Learn: Protecting Multimodal Data from Unauthorized Fine-Tuning of Large Vision-Language Model

Chengshuai Zhao, Dawei Li, Huan Liu, Zhen Tan, Zhiyuan Yu

Data owners can add invisible perturbations to images and text to stop large vision-language models from learning real content during unauthorized fine-tuning.

arxiv:2605.14291 v1 · 2026-05-14 · cs.CR · cs.AI · cs.CL · cs.CV · cs.LG

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{TKNCVVC76J3KRK2NEPNL3GJYPY}

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

MMGuard is evaluated against nine open-source LVLMs across six datasets. Our comprehensive results demonstrate effective, stealthy, and robust protection under white-box, gray-box, and black-box threat models, establishing a mechanistic advantage in proactively defending against aggressive fine-tuning exploitation.

C2weakest assumption

The injected perturbations create an optimization shortcut that forces the model to overfit to noise rather than content, and the cross-modal binding disruption enforces spurious correlations with theoretical guarantees, assuming the LVLM's attention and loss landscape behave predictably under the chosen perturbation strategy.

C3one line summary

MMGuard generates unlearnable multimodal examples via perturbations that exploit LVLM optimization shortcuts and disrupt cross-modal bindings, providing robust protection against unauthorized fine-tuning across threat models.

References

75 extracted · 75 resolved · 5 Pith anchors

[1] IEEE Transactions on Multimedia , year=

[2] Bai, Shuai and Chen, Keqin and Liu, Xuejing and Wang, Jialin and Ge, Wenbin and Song, Sibo and Dang, Kai and Wang, Peng and Wang, Shijie and Tang, Jun and others , journal=

[3] Qwen3-VL Technical Report · arXiv:2511.21631

[4] 2021 IEEE Symposium on Security and Privacy , pages= 2021

[5] 34th USENIX Security Symposium (USENIX Security 25) , pages=

Receipt and verification

First computed	2026-05-17T23:39:10.196919Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

9a9a2ad45ff276a8ab4d23dabd99387e212d2cff4a32ced6f211e825a2efd6c1

Aliases

arxiv: 2605.14291 · arxiv_version: 2605.14291v1 · doi: 10.48550/arxiv.2605.14291 · pith_short_12: TKNCVVC76J3K · pith_short_16: TKNCVVC76J3KRK2N · pith_short_8: TKNCVVC7

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/TKNCVVC76J3KRK2NEPNL3GJYPY \
  | 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: 9a9a2ad45ff276a8ab4d23dabd99387e212d2cff4a32ced6f211e825a2efd6c1

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "c32aac0e42aa4742efa0365c16d87960036e04a988d3cf4df2c932d6f71f6a00",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CL",
      "cs.CV",
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CR",
    "submitted_at": "2026-05-14T02:49:27Z",
    "title_canon_sha256": "29eeeb6ea16a1c46e35bdb94b5ba65d1cda12753ab4858d3691072b88df22a92"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.14291",
    "kind": "arxiv",
    "version": 1
  }
}