Pith Number

pith:CDQRCQ33

pith:2026:CDQRCQ33KF6XF7PQWYRKJHFXYF

not attested not anchored not stored refs resolved

DiRotQ: Rotation-Aware Quantization for 4-bit Diffusion Transformers

Hesham Mostafa, Mahsa Salmani, Sayeh Sharify

DiRotQ rotates activations into a PCA-derived basis to protect dominant variance directions during 4-bit quantization of diffusion transformers.

arxiv:2605.16732 v1 · 2026-05-16 · 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{CDQRCQ33KF6XF7PQWYRKJHFXYF}

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

DiRotQ achieves an FID of 15.9 and PSNR of 19.1 dB on PixArt-Σ over the MJHQ-30K dataset, outperforming the prior state-of-the-art SVDQuant (FID 18.9, PSNR 17.6) under the same INT W4A4 setting.

C2weakest assumption

That a low-rank subspace identified via PCA on calibration data captures the dominant activation variance sufficiently to allow safe 4-bit quantization of the remaining components, with the rotation and fusion preserving overall model behavior.

C3one line summary

DiRotQ uses PCA-based rotation-aware activation quantization combined with GPTQ to achieve better FID and PSNR in 4-bit diffusion transformers than prior methods like SVDQuant.

References

88 extracted · 88 resolved · 16 Pith anchors

[1] QuaRot: Outlier-free 4-bit inference in rotated LLMs 2024

[2] Blended diffusion for text-driven editing of natural images 2022

[3] eDiff-I: Text-to-Image Diffusion Models with an Ensemble of Expert Denoisers 2022 · arXiv:2211.01324

[4] All are worth words: A ViT backbone for diffusion models 2023

[5] Token Merging: Your ViT But Faster 2022 · arXiv:2210.09461

Formal links

1 machine-checked theorem link

Receipt and verification

First computed	2026-05-20T00:02:38.876958Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

10e111437b517d72fdf0b622a49cb7c15baf9c2ebc989483950a99434a7ab821

Aliases

arxiv: 2605.16732 · arxiv_version: 2605.16732v1 · doi: 10.48550/arxiv.2605.16732 · pith_short_12: CDQRCQ33KF6X · pith_short_16: CDQRCQ33KF6XF7PQ · pith_short_8: CDQRCQ33

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/CDQRCQ33KF6XF7PQWYRKJHFXYF \
  | 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: 10e111437b517d72fdf0b622a49cb7c15baf9c2ebc989483950a99434a7ab821

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "20c847605eec2cd556418632f4e5b97b8838594a815ba1e04c381e3efb3630d0",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by-nc-sa/4.0/",
    "primary_cat": "cs.CV",
    "submitted_at": "2026-05-16T00:52:00Z",
    "title_canon_sha256": "a2b476cfaaa63abd6d351caba3f63d3e8f609d1f6bc91b9c028d1819592afa2d"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.16732",
    "kind": "arxiv",
    "version": 1
  }
}