pith:6JPDOEJO
Forget Many, Forget Right: Scalable and Precise Concept Unlearning in Diffusion Models
ScaPre gives diffusion models a closed-form way to unlearn many concepts at once by isolating relevant parameters and stabilizing updates.
arxiv:2601.06162 v4 · 2026-01-06 · cs.LG · cs.CV
Add to your LaTeX paper
\usepackage{pith}
\pithnumber{6JPDOEJOYRA5FEWUO6ZI7PUYDL}
Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge
Record completeness
Claims
ScaPre yields an efficient closed-form solution without requiring auxiliary data or sub-models. It forgets up to ×5 more concepts than the best baseline within acceptable quality limits, achieving state-of-the-art precision and efficiency for large-scale unlearning.
That the Informax Decoupler can accurately identify concept-relevant parameters and strictly confine updates to the target subspace without collateral damage, and that spectral trace regularization plus geometry alignment fully resolve conflicting weight updates across large numbers of concepts.
ScaPre delivers a closed-form method for precise large-scale multi-concept unlearning in diffusion models that removes up to five times more concepts than prior baselines while preserving image quality.
Cited by
Receipt and verification
| First computed | 2026-05-20T00:04:20.754650Z |
|---|---|
| Builder | pith-number-builder-2026-05-17-v1 |
| Signature | Pith Ed25519
(pith-v1-2026-05) · public key |
| Schema | pith-number/v1.0 |
Canonical hash
f25e37112ec441d292d477b28fbe981adf8a7d1ee6b2778c7e2f73f6bfc0a106
Aliases
· · · · ·Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/6JPDOEJOYRA5FEWUO6ZI7PUYDL \
| 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: f25e37112ec441d292d477b28fbe981adf8a7d1ee6b2778c7e2f73f6bfc0a106
Canonical record JSON
{
"metadata": {
"abstract_canon_sha256": "1a465136b825769debdf97ba9de6b47e9d0cbfec3dd49e71583602ab791d87f9",
"cross_cats_sorted": [
"cs.CV"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"primary_cat": "cs.LG",
"submitted_at": "2026-01-06T23:59:17Z",
"title_canon_sha256": "b85f83a57e5451235104f6799118ead66deada61c9c148502c67ba2381756b11"
},
"schema_version": "1.0",
"source": {
"id": "2601.06162",
"kind": "arxiv",
"version": 4
}
}