Pith. sign in

REVIEW 5 major objections 5 minor 44 references

Pack-PTQ: Advancing Post-training Quantization of Neural Networks by Pack-wise Reconstruction

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Pack-PTQ claims that quantizing networks in Hessian-guided packs, rather than block by block, keeps low-bit models close to full-precision accuracy.

desk verdict Pack-wise reconstruction is a genuine idea, but the paper's headline results rest on implausible baseline numbers and a shaky Hessian approximation; fixable, but not credible as written. read the letter →

arxiv 2505.00259 v1 pith:WT4SXPD2 submitted 2025-05-01 cs.CV cs.AI

Changjun Li , Runqing Jiang , Zhuo Song , Pengpeng Yu , Ye Zhang , Yulan Guo This is my paper · ORCID
classification cs.CVcs.AI
keywords post-trainingquantizationcross-blockdependencyHessian-guidedpackingmixed-precisionlow-bitvisiontransformerspointcloudclassificationblockreconstruction
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the standard reconstruction unit in post-training quantization—one block of layers—is the wrong granularity for low-bit compression. Its proposal, Pack-PTQ, scores each block with a Hessian-derived importance measure, groups consecutive blocks into non-overlapping packs, and reconstructs each pack so that its quantized output matches the full-precision output, thereby preserving dependencies between blocks. A second component assigns different bit-widths to different packs under a memory constraint. The paper reports ImageNet results across CNNs and vision transformers, including ViT-S at W3/A3 rising from 0.42% with block-wise reconstruction to 55.56% with packing and mixed precision, and near-full-precision accuracy on ModelNet40 point cloud classification with PointNet. The payoff, if the claim holds, is accurate 3-bit and 4-bit deployment without retraining, using only a small calibration set.

What carries the argument

The load-bearing object is the Hessian-guided block importance score $S = \mu_H(z) \approx \mathbb{E}[2(L_q^b - \Delta z^\top g(z))]\,/\,\mathbb{E}[\Delta z^\top \Delta z]$, derived from a second-order Taylor expansion of the loss change when a block's output $z$ is perturbed by quantization noise $\Delta z$. Theorem 1 justifies reading $\mu_H(z)$ as the mean of all entries of the Hessian $H(z)$ by assuming i.i.d. Gaussian $\Delta z$ and approximating $\operatorname{tr}(H(z)\sigma^2 I)$ by $\mu_H(z)\sigma^2 n$. Starting from the last block, the packing algorithm finds the lowest-scoring block in the remaining range and makes a pack from that block to the end, then repeats; this yields non-overlapping packs as the reconstruction unit. The second mechanism is pack-level mixed precision: bit-widths are chosen to maximize $\sum_j b_j\Omega_j$ under a memory budget, where $\Omega_j$ averages block sensitivities and quantization losses in pack $j$. These two mechanisms are what the paper credits for capturing cross-block dependency while keeping the reconstruction tractable.

What would settle it

On ResNet18 at W3/A3, compute each block's importance score with exact Hessian-vector products instead of the Gaussian approximation, run the same packing and reconstruction, and compare top-1 accuracy with the reported 64.46% without mixed precision; if exact-Hessian packing does not reproduce or exceed that number, the approximation, not pack-wise reconstruction itself, is what matters.

Watch

Extended reading notes

Core claim

The central claim is that cross-block dependency, not just per-block sensitivity, limits low-bit post-training quantization, and that an adaptive packing mechanism can preserve that dependency without reconstructing the entire network. Pack-PTQ computes a Hessian-guided importance score for each block, then repeatedly selects the lowest-scoring block in the not-yet-packed range and groups it with all later blocks, so every pack begins with a relatively insensitive block and contains the blocks that follow it. The paper argues that jointly reconstructing such packs—aligning quantized pack outputs with full-precision pack outputs—yields more accurate quantization parameters than block-wise reconstruction. On top of this, pack-level mixed precision solves a constrained optimization that gives more bits to high-sensitivity packs. The reported experiments claim consistent gains at W3/A3 and W4/A4 on ImageNet CNN and vision transformer models, with especially large margins on transformers, and stable near-full-precision results on ModelNet40.

Load-bearing premise

The packing mechanism rests on the approximation that a block's Hessian can be summarized by its average entry, computed as if quantization noise were independent Gaussian; real quantization noise is not Gaussian and general Hessians need not satisfy that equality, so the scores that decide pack boundaries are only as sound as that shortcut.

Editorial extensions

If this is right

  • If the central claim is right, the reconstruction granularity for PTQ shifts from fixed single blocks to adaptive, variable-size packs, so methods no longer need to choose between block-wise efficiency and network-wise dependency capture.
  • Low-bit vision transformers become practical: the paper reports W3/A3 ViT accuracy far above prior PTQ baselines, for example DeiT-S at 58.38% with mixed precision versus 29.33% for PD-Quant.
  • Pack-level mixed precision gives a principled way to spend a memory budget: high-sensitivity packs get more bits, low-sensitivity packs get fewer, under an explicit constraint.
  • The same recipe transfers to a new domain, 3D point cloud classification, where the paper reports W3/A3 accuracy within about half a point of full precision and even above it with mixed precision.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The authors do not isolate whether the Hessian approximation or the optimization effect of jointly reconstructing more layers drives the gains; a version using exact Hessian-vector products, or scores replaced by random ones under the same pack sizes, would separate the two.
  • Because every pack starts at a low-importance block and includes all later blocks up to the previous pack boundary, the method may implicitly place quantization 'buffers' at pack boundaries; the paper does not test whether boundary position, rather than internal dependency, explains part of the benefit.
  • The packing idea should extend naturally to detection or segmentation backbones and to large language models, where cross-layer dependencies are also strong; that is an extrapolation, not a claim in the paper.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 5 minor

Summary. This manuscript proposes Pack-PTQ, a post-training quantization method. It computes per-block Hessian-guided importance scores, partitions blocks into contiguous non-overlapping packs, reconstructs the network pack-wise, and assigns mixed precisions to packs. The paper reports experiments on ImageNet with CNN and ViT architectures and on ModelNet40 with PointNet, claiming consistent superiority over existing PTQ methods. The core novelty is the packing granularity and the Hessian-guided scoring; the paper also includes an ablation study and an execution-time analysis.

Significance. If the results were valid, pack-wise reconstruction would be a useful contribution to post-training quantization, and the combination of Hessian-guided packing with mixed precision could improve low-bit performance across architectures. The paper's coverage of CNNs, ViTs, and point-cloud networks, as well as its inclusion of ablations, is appropriate. However, the central empirical claims are not credible as reported: the baseline scores in Table 2 are far below published results and at or below chance level in several cases, and the point-cloud results in Table 3 show quantized models matching or exceeding full-precision accuracy at 2-bit weights without explanation. The theoretical derivation supporting the packing criterion is also mathematically incorrect. These issues affect load-bearing parts of the contribution.

major comments (5)
  1. [Table 2, W3/A3 rows] Table 2 reports near-zero W3/A3 accuracies for baselines, e.g., PTQ4ViT at 0.01% on ViT-B and DeiT-S, RepQ-ViT at 0.14% on ViT-B, and BRECQ at 0.42% on ViT-S. These are at or below the 0.1% random-guess level for ImageNet and are orders of magnitude below the same methods' W4/A4 results in the same table (e.g., RepQ-ViT ViT-B reaches 68.48% at W4/A4). A one-bit reduction does not plausibly cause such a collapse. Because the paper's headline claim of superiority is based on comparisons to these baselines, the empirical comparison is not credible as reported.
  2. [Table 3] Table 3 reports Pack-PTQ at W3/A3 with mixed precision (92.28 mAcc, 89.53 OA) and at W4/A4 with mixed precision (92.21 mAcc, 88.56 OA), both at or above the stated full-precision PointNet accuracy of 92.01 mAcc and 88.54 OA. Even W2/A4 without mixed precision loses only 0.45 mAcc and 0.17 OA. Quantization that improves on full precision, especially at 2-bit weights, requires an explanation such as a regularizing effect or an evaluation artifact, but none is provided; this makes the point-cloud evidence difficult to credit.
  3. [Theorem 1, Eq. (5)] Equation (5) asserts tr(H(z)σ²I) ≈ μ_H(z)σ²n, but tr(H(z)σ²I)=σ²Σ_i H_ii(z); equating this to μ_H(z)σ²n requires the mean of the diagonal entries to equal the mean of all entries of H, which is not true for a general Hessian. The proof also assumes Δz has i.i.d. Gaussian components, whereas quantization noise is deterministic and input-dependent. Since Eq. (8) uses this approximation to define the importance score S that drives the adaptive packing, the theoretical grounding of the packing mechanism is invalid as stated.
  4. [Section 4.2, Eqs. (10)-(11)] Equation (11) defines the mixed-precision objective in terms of L_q^j[i], the quantization loss of the i-th block in pack j, but the bit-width assignment problem is solved before the pack reconstruction is performed. The paper does not specify how L_q^j[i] is computed for candidate bit-widths or whether it depends on the very quantizers being optimized, which creates a circularity concern. The memory constraint C and the candidate bit-width set K used in the experiments are also not reported, so the mixed-precision results cannot be interpreted as a constrained optimization.
  5. [Abstract and Section 5.2] The abstract and Section 5.2 state that Pack-PTQ consistently outperforms existing methods in both W3/A3 and W4/A4 settings, but Table 2 shows Adalog outperforming Pack-PTQ at W4/A4 on ViT-S (72.13 vs 61.03 w/o MP and 66.83 with MP) and ViT-B (79.16 vs 75.71 w/o MP and 77.89 with MP). The claim should be qualified to match the reported table.
minor comments (5)
  1. [Section 5.3] There are typographical and grammatical errors in this section, including 'quantitation parameters' and 'the block-wise overlooks the cross-block relationships'; these should be corrected.
  2. [Section 6] The section heading 'Conculsion' is misspelled and should read 'Conclusion'.
  3. [References] References [19] and [22] are the same QDrop paper listed twice, and the citation formatting is inconsistent between the two entries.
  4. [Section 5.1.2] The source of the pre-trained PointNet model and the exact calibration protocol for ModelNet40 should be stated; the reported full-precision accuracy of 92.01 mAcc and 88.54 OA should also be reconciled with commonly reported PointNet accuracy.
  5. [Figure 1] The values in the bar chart are difficult to read in the current figure; numerical labels or a larger figure would improve clarity.

Circularity Check

2 steps flagged · score 6.0 of 10

Theorem 1 proves its Hessian-mean approximation by assuming it, and the mixed-precision objective scores bit-widths using the very quantization loss that depends on the chosen bit-width; the packing ablation offers partial independent empirical support.

  1. other [Section 4.1, Theorem 1 and Eqs. (5)-(8)]
    "E[∆z⊤H(z)∆z] = tr(H(z)E[∆z∆z⊤]) = tr(H(z)σ2I)≈µH(z)σ2n. ... µH(z) = µH(z)σ2n / σ2n ≈ E[∆z⊤H(z)∆z] / E[∆z⊤∆z]."

    Theorem 1 claims to prove that the all-element Hessian mean µ_H is approximated by the quadratic-form ratio, which under i.i.d. Gaussian noise equals the average diagonal of H. But Eq. (5) already replaces tr(H σ² I) with µ_H σ² n, which is exactly the assertion that the average diagonal equals the all-element mean — the very statement the theorem is supposed to establish. Eq. (7) then divides by σ² n and recovers the conclusion from that assumed equality. The approximation is thus postulated, not derived. Since Eq. (8) sets the block importance score S = µ_H(z), the 'Hessian-guided' packing criterion rests on this circularly proved quantity.

  2. self definitional [Section 4.2, Eqs. (10)-(11)]
    "max {b1,b2,...,bM} ΣM j=1 bj· Ωj, s.t. ΣM j=1 bj·pj≤C ... Ωj = 1/nj Σnj i=1 (Sj[i]·L q j[i]) ... Lq j[i] denotes the quantization loss of the i-th block in the j-th pack."

    The objective coefficient Ω_j is supposed to measure pack sensitivity so that Eq. (10) can choose the bit-width b_j for that pack. But Ω_j is defined through L_q^j[i], the quantization loss of the block, and quantization loss is inherently a function of the bit-width actually used. As written, the coefficient of the decision variable b_j therefore already depends on the value of b_j being selected. Absent an explicit statement that L_q is evaluated at a fixed reference bit-width independent of b_j, the mixed-precision assignment reduces to choosing a bit-width from a quantity that presupposes that bit-width, making the objective self-referential.

full rationale

Two genuinely circular/self-referential steps were found in the method's derivation. Theorem 1 (Sec. 4.1, Eqs. 5-7) derives the Hessian-mean approximation by assuming the key equality tr(H σ² I) ≈ µ_H σ² n, which is exactly the statement being proved; the importance score S = µ_H in Eq. (8) inherits this unproven assumption. Separately, the mixed-precision objective (Sec. 4.2, Eqs. 10-11) defines pack sensitivity Ω_j using L_q^j[i], the block quantization loss, which is a function of the bit-width b_j being optimized, making the objective coefficient self-referential as written. No load-bearing self-citation chain exists, and the packing-vs-no-packing ablation in Table 4 provides empirical support for grouping blocks that is independent of the circular proof. The implausible baseline numbers in Table 2 are an empirical correctness/protocol concern rather than circularity and were not scored. Overall, the theory behind the headline mechanism is partially circular, but the empirical packing effect has independent content, warranting a score of 6.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central mechanism rests on a Hessian approximation that is not proven, an unspecified initial quantizer to compute the scores, and an unreported memory budget for mixed precision. The ledger shows that the 'Hessian-guided' part is more of a heuristic than a derived quantity.

free parameters (3)
  • Memory constraint C for mixed-precision assignment = Not reported
    In Eq. (10), C sets the total bit budget that determines the bit-width for each pack. The paper never reports the value of C or how it maps to the reported 'with MP' results, so the bit assignments are not reproducible.
  • Candidate bit-width set K = Not reported
    The set K of allowed bit-widths in Eq. (10) is not specified in the experiments (e.g., for W3/A3 with MP, is it {2,3,4} or {3,4}?). The results depend on this choice.
  • Initial quantization for computing block scores L_q^b = Not reported
    The block importance score S in Eq. (8) requires L_q^b, the quantization loss of a block, which depends on initial quantization parameters. The paper does not specify what initial quantizer is used before packing, so the scores (and hence packs) are not uniquely determined.
assumptions (5)
  • ad hoc to paper tr(H σ² I) ≈ μ_H σ² n in the proof of Theorem 1
    This equality holds only when the average Hessian diagonal equals the average of all Hessian entries, which is not generally true and is not stated as an assumption.
  • ad hoc to paper Δz has i.i.d. N(0, σ²) components
    The theorem's proof relies on this Gaussian i.i.d. structure, but quantization noise is deterministic given the input and weights, and elements are correlated. This is an unrealistic modeling assumption.
  • domain assumption The loss degradation is well approximated by a second-order Taylor expansion with gradient term (Eq. 2)
    This is a standard PTQ assumption (also used in BRECQ), but it is an approximation that can fail at very low bit widths.
  • ad hoc to paper Contiguous packs capture the cross-block dependencies that matter
    The packing algorithm only groups consecutive blocks (Eq. 9). Dependencies from residual connections or attention across non-adjacent blocks are ignored. The paper provides no evidence that non-local dependencies are negligible.
  • ad hoc to paper The importance score S indicates the influence of a block's output on 'preceding blocks' (Section 4.1)
    A block's output influences subsequent blocks, not preceding ones; the direction of causality in the stated metric is unclear and appears to be a conceptual error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pack-PTQ: Advancing Post-training Quantization of Neural Networks by Pack-wise Reconstruction." pith.science (2026). https://pith.science/paper/WT4SXPD2

@misc{pith2026250500259,
  author       = {Pith},
  title        = {Pith review of: Pack-PTQ: Advancing Post-training Quantization of Neural Networks by Pack-wise Reconstruction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WT4SXPD2}},
  note         = {Machine review of arXiv:2505.00259}
}
read the original abstract

Post-training quantization (PTQ) has evolved as a prominent solution for compressing complex models, which advocates a small calibration dataset and avoids end-to-end retraining. However, most existing PTQ methods employ block-wise reconstruction, which neglects cross-block dependency and exhibits a notable accuracy drop in low-bit cases. To address these limitations, this paper presents a novel PTQ method, dubbed Pack-PTQ. First, we design a Hessian-guided adaptive packing mechanism to partition blocks into non-overlapping packs, which serve as the base unit for reconstruction, thereby preserving the cross-block dependency and enabling accurate quantization parameters estimation. Second, based on the pack configuration, we propose a mixed-precision quantization approach to assign varied bit-widths to packs according to their distinct sensitivities, thereby further enhancing performance. Extensive experiments on 2D image and 3D point cloud classification tasks, using various network architectures, demonstrate the superiority of our method over the state-of-the-art PTQ methods.

Figures

Figures reproduced from arXiv: 2505.00259 by the authors.

Figure 1
Figure 1. Quantization results of different reconstruction strategies on ImageNet with W3/A3 setting. “No Packing” [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overview of our proposed Pack-PTQ method. We begin by computing individual block scores, which take [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization of important scores of blocks on (a) ResNet18 and (b) ViT-S. Different colors denote different [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 37 canonical work pages

  1. [1]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020

    arXiv 2010

  2. [2]

    Training data-efficient image transformers & distillation through attention

    Hugo Touvron, Matthieu Cord, Matthijs Douze, Francisco Massa, Alexandre Sablayrolles, and Hervé Jégou. Training data-efficient image transformers & distillation through attention. In International Conference on Machine Learning, pages 10347–10357, 2021

    work page 2021

  3. [3]

    Swin transformer: Hierarchical vision transformer using shifted windows

    Ze Liu, Yutong Lin, Yue Cao, Han Hu, Yixuan Wei, Zheng Zhang, Stephen Lin, and Baining Guo. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE International Conference on Computer Vision, pages 10012–10022, 2021

    work page 2021

  4. [4]

    Mask R-CNN

    Kaiming He, Georgia Gkioxari, Piotr Dollár, and Ross Girshick. Mask R-CNN. In Proceedings of the IEEE International Conference on Computer Vision, pages 2961–2969, 2017

    work page 2017

  5. [5]

    Dual-mode learning for multi-dataset x-ray security image detection

    Fenghong Yang, Runqing Jiang, Yan Yan, Jing-Hao Xue, Biao Wang, and Hanzi Wang. Dual-mode learning for multi-dataset x-ray security image detection. IEEE Transactions on Information F orensics and Security, 2024

    work page 2024

  6. [6]

    Segmenter: Transformer for semantic segmentation

    Robin Strudel, Ricardo Garcia, Ivan Laptev, and Cordelia Schmid. Segmenter: Transformer for semantic segmentation. In Proceedings of the IEEE International Conference on Computer Vision , pages 7262–7272, 2021

    work page 2021

  7. [7]

    Segformer: Simple and efficient design for semantic segmentation with transformers

    Enze Xie, Wenhai Wang, Zhiding Yu, Anima Anandkumar, Jose M Alvarez, and Ping Luo. Segformer: Simple and efficient design for semantic segmentation with transformers. In Advances in Neural Information Processing Systems, pages 12077–12090, 2021

    work page 2021

  8. [8]

    MobileNetv2: Inverted residuals and linear bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zhmoginov, and Liang-Chieh Chen. MobileNetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 4510–4520, 2018

    work page 2018

Show all 44 references
  1. [9]

    A survey on deep neural network pruning: Taxonomy, comparison, analysis, and recommendations

    Hongrong Cheng, Miao Zhang, and Javen Qinfeng Shi. A survey on deep neural network pruning: Taxonomy, comparison, analysis, and recommendations. IEEE Transactions on Pattern Analysis and Machine Intelligence , 46:10558–10578, 2024

    2024

  2. [10]

    When sparse neural network meets label noise learning: A multistage learning framework

    Runqing Jiang, Yan Yan, Jing-Hao Xue, Biao Wang, and Hanzi Wang. When sparse neural network meets label noise learning: A multistage learning framework. IEEE Transactions on Neural Networks and Learning Systems , 35(2):2208–2222, 2022

    2022

  3. [11]

    Knowledge distillation: A survey

    Jianping Gou, Baosheng Yu, Stephen J Maybank, and Dacheng Tao. Knowledge distillation: A survey. Interna- tional Journal of Computer Vision, 129(6):1789–1819, 2021

    2021

  4. [12]

    Knowledge distillation meets label noise learning: Ambiguity-guided mutual label refinery

    Runqing Jiang, Yan Yan, Jing-Hao Xue, Si Chen, Nannan Wang, and Hanzi Wang. Knowledge distillation meets label noise learning: Ambiguity-guided mutual label refinery. IEEE Transactions on Neural Networks and Learning Systems, pages 1–14, 2023

    2023

  5. [13]

    A white paper on neural network quantization

    Markus Nagel, Marios Fournarakis, Rana Ali Amjad, Yelysei Bondarenko, Mart Van Baalen, and Tijmen Blankevoort. A white paper on neural network quantization. arXiv preprint arXiv:2106.08295, 2021

    2021 arXiv

  6. [14]

    BRECQ: Pushing the limit of post-training quantization by block reconstruction

    Yuhang Li, Ruihao Gong, Xu Tan, Yang Yang, Peng Hu, Qi Zhang, Fengwei Yu, Wei Wang, and Shi Gu. BRECQ: Pushing the limit of post-training quantization by block reconstruction. In International Conference on Learning Representations, 2021

    2021

  7. [15]

    Quantization and training of neural networks for efficient integer-arithmetic-only inference

    Benoit Jacob, Skirmantas Kligys, Bo Chen, Menglong Zhu, Matthew Tang, Andrew Howard, Hartwig Adam, and Dmitry Kalenichenko. Quantization and training of neural networks for efficient integer-arithmetic-only inference. In Proceedings of the IEEE Conference on Computer Vision an...

    2018

  8. [16]

    PACT: Parameterized clipping activation for quantized neural networks

    Jungwook Choi, Zhuo Wang, Swagath Venkataramani, Pierce I-Jen Chuang, Vijayalakshmi Srinivasan, and Kailash Gopalakrishnan. PACT: Parameterized clipping activation for quantized neural networks. arXiv preprint arXiv:1805.06085, 2018

    2018 arXiv

  9. [17]

    PTQ4ViT: Post-training quantization for vision transformers with twin uniform quantization

    Zhihang Yuan, Chenhao Xue, Yiqi Chen, Qiang Wu, and Guangyu Sun. PTQ4ViT: Post-training quantization for vision transformers with twin uniform quantization. In European Conference on Computer Vision, pages 191–207, 2022. 10 Running Title for Header

    2022

  10. [18]

    Up or down? adaptive rounding for post-training quantization

    Markus Nagel, Rana Ali Amjad, Mart Van Baalen, Christos Louizos, and Tijmen Blankevoort. Up or down? adaptive rounding for post-training quantization. In International Conference on Machine Learning , pages 7197–7206, 2020

    2020

  11. [19]

    Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization

    Xiuying Wei, Ruihao Gong, Yuhang Li, Xianglong Liu, and Fengwei Yu. Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization. arXiv preprint arXiv:2203.05740, 2022

    2022 arXiv

  12. [20]

    Learned step size quantization

    Steven K Esser, Jeffrey L McKinstry, Deepika Bablani, Rathinakumar Appuswamy, and Dharmendra S Modha. Learned step size quantization. arXiv preprint arXiv:1902.08153, 2019

    1902 arXiv

  13. [21]

    Overcoming oscillations in quantization-aware training

    Markus Nagel, Marios Fournarakis, Yelysei Bondarenko, and Tijmen Blankevoort. Overcoming oscillations in quantization-aware training. In International Conference on Machine Learning , pages 16318–16330, 2022

    2022

  14. [22]

    Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization, 2023

    Xiuying Wei, Ruihao Gong, Yuhang Li, Xianglong Liu, and Fengwei Yu. Qdrop: Randomly dropping quantization for extremely low-bit post-training quantization, 2023

    2023

  15. [23]

    NoisyQuant: Noisy bias- enhanced post-training activation quantization for vision transformers

    Yijiang Liu, Huanrui Yang, Zhen Dong, Kurt Keutzer, Li Du, and Shanghang Zhang. NoisyQuant: Noisy bias- enhanced post-training activation quantization for vision transformers. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 20321–20330, 2023

    2023

  16. [24]

    Lightweight maize disease detection through post-training quantization with similarity preservation

    Carlos Victorino Padeiro, Tse-Wei Chen, Takahiro Komamizu, and Ichiro Ide. Lightweight maize disease detection through post-training quantization with similarity preservation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2111–2120, 2024

    2024

  17. [25]

    PD-Quant: Post-training quantization based on prediction difference metric

    Jiawei Liu, Lin Niu, Zhihang Yuan, Dawei Yang, Xinggang Wang, and Wenyu Liu. PD-Quant: Post-training quantization based on prediction difference metric. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 24427–24437, 2023

    2023

  18. [26]

    Data-free quantization through weight equalization and bias correction

    Markus Nagel, Mart van Baalen, Tijmen Blankevoort, and Max Welling. Data-free quantization through weight equalization and bias correction. In Proceedings of the IEEE International Conference on Computer Vision , pages 1325–1334, 2019

    2019

  19. [27]

    Towards mixed-precision quantization of neural networks via constrained optimization

    Weihan Chen, Peisong Wang, and Jian Cheng. Towards mixed-precision quantization of neural networks via constrained optimization. In Proceedings of the IEEE International Conference on Computer Vision , pages 5350–5359, 2021

    2021

  20. [28]

    HAWQ: Hessian aware quantization of neural networks with mixed-precision

    Zhen Dong, Zhewei Yao, Amir Gholami, Michael W Mahoney, and Kurt Keutzer. HAWQ: Hessian aware quantization of neural networks with mixed-precision. In Proceedings of the IEEE International Conference on Computer Vision, pages 293–302, 2019

    2019

  21. [29]

    HAWQ-v2: Hessian aware trace-weighted quantization of neural networks

    Zhen Dong, Zhewei Yao, Daiyaan Arfeen, Amir Gholami, Michael W Mahoney, and Kurt Keutzer. HAWQ-v2: Hessian aware trace-weighted quantization of neural networks. In Advances in Neural Information Processing Systems, pages 18518–18529, 2020

    2020

  22. [30]

    APTQ: Attention-aware post- training mixed-precision quantization for large language models

    Ziyi Guan, Hantao Huang, Yupeng Su, Hong Huang, Ngai Wong, and Hao Yu. APTQ: Attention-aware post- training mixed-precision quantization for large language models. In Proceedings of the ACM/IEEE Design Automation Conference, pages 1–6, 2024

    2024

  23. [31]

    ImageNet classification with deep convolutional neural networks

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. ImageNet classification with deep convolutional neural networks. pages 1106–1114, 2012

    2012

  24. [32]

    RapQ: Rescuing accuracy for power-of-two low-bit post-training quantization

    Hongyi Yao, Pu Li, Jian Cao, Xiangcheng Liu, Chenying Xie, and Bingzhang Wang. RapQ: Rescuing accuracy for power-of-two low-bit post-training quantization. arXiv preprint arXiv:2204.12322, 2022

    2022 arXiv

  25. [33]

    Solving oscillation problem in post-training quantization through a theoretical perspective

    Yuexiao Ma, Huixia Li, Xiawu Zheng, Xuefeng Xiao, Rui Wang, Shilei Wen, Xin Pan, Fei Chao, and Rongrong Ji. Solving oscillation problem in post-training quantization through a theoretical perspective. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni...

    2023

  26. [34]

    Genie: Show me the data for quantization

    Yongkweon Jeon, Chungman Lee, and Ho-young Kim. Genie: Show me the data for quantization. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 12064–12073, 2023

    2023

  27. [35]

    Repq-ViT: Scale reparameterization for post-training quantization of vision transformers

    Zhikai Li, Junrui Xiao, Lianwei Yang, and Qingyi Gu. Repq-ViT: Scale reparameterization for post-training quantization of vision transformers. In Proceedings of the IEEE International Conference on Computer Vision , pages 17227–17236, 2023

    2023

  28. [36]

    3d shapenets: A deep representation for volumetric shapes

    Zhirong Wu, Shuran Song, Aditya Khosla, Fisher Yu, Linguang Zhang, Xiaoou Tang, and Jianxiong Xiao. 3d shapenets: A deep representation for volumetric shapes. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 1912–1920, 2015

    1912

  29. [37]

    I&S-ViT: An inclusive & stable method for pushing the limit of post-training vits quantization

    Yunshan Zhong, Jiawei Hu, Mingbao Lin, Mengzhao Chen, and Rongrong Ji. I&S-ViT: An inclusive & stable method for pushing the limit of post-training vits quantization. arXiv preprint arXiv:2311.10126, 2023. 11 Running Title for Header

    2023

  30. [38]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 770–778, 2016

    2016

  31. [39]

    Designing network design spaces

    Ilija Radosavovic, Raj Prateek Kosaraju, Ross Girshick, Kaiming He, and Piotr Dollár. Designing network design spaces. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition , pages 10428–10436, 2020

    2020

  32. [40]

    Mnasnet: Platform-aware neural architecture search for mobile

    Mingxing Tan, Bo Chen, Ruoming Pang, Vijay Vasudevan, Mark Sandler, Andrew Howard, and Quoc V Le. Mnasnet: Platform-aware neural architecture search for mobile. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 2820–2828, 2019

    2019

  33. [41]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 652–660, 2017

    2017

  34. [42]

    Texq: Zero-shot network quantization with texture feature distribution calibration

    Xinrui Chen, Yizhi Wang, Renao Yan, Yiqing Liu, Tian Guan, and Yonghong He. Texq: Zero-shot network quantization with texture feature distribution calibration. In Advances in Neural Information Processing Systems , 2024

    2024

  35. [43]

    AdaLog: Post-training quantization for vision transformers with adaptive logarithm quantizer

    Zhuguanyu Wu, Jiaxin Chen, Hanwen Zhong, Di Huang, and Yunhong Wang. AdaLog: Post-training quantization for vision transformers with adaptive logarithm quantizer. In European Conference on Computer Vision, pages 411–427, 2025

    2025

  36. [44]

    Towards accu- rate post-training quantization for vision transformer

    Yifu Ding, Haotong Qin, Qinghua Yan, Zhenhua Chai, Junjie Liu, Xiaolin Wei, and Xianglong Liu. Towards accu- rate post-training quantization for vision transformer. In Proceedings of the 30th ACM International Conference on Multimedia, pages 5380–5388, 2022. 12

    2022

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.