Pith Number

pith:MMLQUVCX

pith:2025:MMLQUVCXA32J35G5AJ2JE6TDDK

not attested not anchored not stored refs resolved

RePack then Refine: Efficient Diffusion Transformer with Vision Foundation Model

Chao Gao, Guanfang Dong, Luke Schultz, Negar Hassanpour

Compressing VFM features to a low-dimensional manifold lets DiTs reach FID 1.82 on ImageNet in 64 epochs, then a refiner improves it to 1.65.

arxiv:2512.12083 v3 · 2025-12-12 · cs.CV

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

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{MMLQUVCXA32J35G5AJ2JE6TDDK}

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

On ImageNet-1K, RePack-DiT-XL/1 achieves an FID of 1.82 in only 64 training epochs. With the Refiner module, performance further improves to an FID of 1.65, significantly surpassing latest LDMs in terms of convergence efficiency.

C2weakest assumption

That the RePack projection to a compact manifold preserves essential structural information from VFM features such that the subsequent Latent-Guided Refiner can reliably restore high-frequency details without introducing artifacts or requiring extensive additional training.

C3one line summary

RePack projects VFM features to a low-dimensional manifold for efficient DiT training, followed by a Latent-Guided Refiner that improves FID to 1.65 on ImageNet-1K after 64 epochs.

References

8 extracted · 8 resolved · 0 Pith anchors

[1] 1936 , month = sep, publisher = 2023 · doi:10.1007/bf02288367

[2] In specific restoration domains such as blind face restoration, methods often leverage strong external priors 2025

[3] II (Compression):We follow the common view that compressing the latent manifold helps simplify the modeling task

[4] III (Semantic Fidelity):We observe that directly utilizing signals from frozen VFMs preserves higher semantic puritycompared to distilling them into a complex learnable encoder

[5] IV (Decoupling):We follow a frequency-decoupled design 1936

Formal links

2 machine-checked theorem links

Cited by

1 paper in Pith

What Matters for Diffusion-Friendly Latent Manifold? Prior-Aligned Autoencoders for Latent Diffusion

Receipt and verification

First computed	2026-05-17T23:39:16.861851Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

63170a545706f49df4dd0274927a631abed14e8fa86ae28ac5628f4917a3dc5e

Aliases

arxiv: 2512.12083 · arxiv_version: 2512.12083v3 · doi: 10.48550/arxiv.2512.12083 · pith_short_12: MMLQUVCXA32J · pith_short_16: MMLQUVCXA32J35G5 · pith_short_8: MMLQUVCX

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/MMLQUVCXA32J35G5AJ2JE6TDDK \
  | 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: 63170a545706f49df4dd0274927a631abed14e8fa86ae28ac5628f4917a3dc5e

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "4d115b9aff5639ffffaf7dc8c8ebe275a1cbdc58caec78d0b510bfb5c891269e",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/publicdomain/zero/1.0/",
    "primary_cat": "cs.CV",
    "submitted_at": "2025-12-12T23:17:41Z",
    "title_canon_sha256": "0645f829babe1a72977dd2b28277921e7c0b124ae6a1f07e5ef5f173b8f2c218"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2512.12083",
    "kind": "arxiv",
    "version": 3
  }
}