Pith. sign in

REVIEW 3 major objections 4 minor 4 cited by

Feature Coding in the Era of Large Models: Dataset, Test Conditions, and Benchmark

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper introduces a public dataset, unified test conditions, and baseline codecs to launch feature coding for large models.

desk verdict First large-model feature coding benchmark; useful dataset and BPFP metric, but the 'unified test conditions' are undercut by baseline-specific truncation ranges. read the letter →

arxiv 2412.04307 v4 pith:UQ2SUXCF submitted 2024-12-05 cs.MM

Changsheng Gao , Yifan Ma , Qiaoxi Chen , Yenan Xu , Dong Liu , Weisi Lin This is my paper · ORCID
classification cs.MM
keywords featurecodinglargemodelsbenchmarktestdatasetBPFPDINOv2Llama3StableDiffusion3
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

Large models are increasingly split across cloud and edge, and the feature tensors passed between segments must be compressed to save bandwidth and storage. This paper argues that existing feature-coding research, built on small convolutional networks and visual data, does not transfer to Transformer-based large models, and that the field lacks the shared resources needed to develop. To fill that gap, it constructs a public dataset of features from DINOv2, Llama3, and Stable Diffusion 3, covering five tasks and both visual and textual modalities, and it fixes unified test conditions including a new bitrate metric, bits per feature point (BPFP). It then evaluates two adapted image codecs on this testbed, showing substantial bitrate savings but visible accuracy loss at low rates and weak correlation between feature distortion and task accuracy for text. The paper's central claim is that large-model feature coding deserves its own research program, one that can now be pursued against a common benchmark.

What carries the argument

The load-bearing object is the proposed testbed: five feature classes extracted at fixed split points of three large models, namely FCls and FSeg from the 40th DINOv2 ViT block, FDpt as a concatenation of the 10th, 20th, 30th, and 40th blocks, FCSR from the 32nd Llama3 decoder layer, and FTTI from the Stable Diffusion 3 VAE decoder input. The evaluation protocol centers on BPFP, bits per feature point, which counts coding bits on the feature tensor itself rather than on source pixels, making bitrate comparable across visual, textual, and generative features. The baseline pipeline is pre-processing (truncation to a central range, 10-bit uniform quantization, packing into a 2D YUV-400 frame), a core codec (VTM intra coding or Hyperprior), and post-processing (unpacking and de-quantization). The fixed split points do the argumentative work: they define what counts as the feature in a distributed large-model system, so the dataset's representativeness rests entirely on those choices.

What would settle it

Extract features from other split points of the same models, for example DINOv2's 10th block or Llama3's 8th decoder layer, run the two baseline codecs, and compare the value distributions and rate-accuracy curves with Table 3 and Figure 2 of the paper; a large mismatch would show the dataset is not representative of large-model features in general.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the features large models exchange have statistics that existing feature-coding benchmarks do not capture. The 40th-block DINOv2 features, 32nd-layer Llama3 features, and Stable Diffusion 3 VAE-decoder-input features show wider, more asymmetric value ranges, sparser and more concentrated distributions, and redundancy patterns (strong vertical correlation for vision transformers, weak spatial correlation for text) that differ from ResNet and Mask R-CNN features used in earlier work. Under two baseline codecs, a handcrafted VVC Intra pipeline and a learned Hyperprior pipeline, these features can be compressed to a fraction of their original 32-bit size, but accuracy drops sharply below a bitrate threshold, and feature mean squared error is not a reliable predictor of task accuracy except for text-to-image synthesis. The intended conclusion is that large-model feature coding is a distinct problem: its data, metrics, and codecs need to be designed specifically for these representations, and the dataset plus test conditions provided here are offered as the common ground for doing so.

Load-bearing premise

The whole testbed rests on the assumption that features extracted at the chosen split points, the 40th DINOv2 block, the 32nd Llama3 decoder layer, and the SD3 VAE decoder input, are representative of what real distributed systems will exchange.

Editorial extensions

If this is right

  • Future feature-coding papers can report results on the same five feature classes, making rate-accuracy numbers directly comparable across methods and papers.
  • BPFP replaces bits-per-pixel as the natural bitrate unit for feature coding, including non-visual features where pixels do not exist.
  • The benchmark shows that off-the-shelf image codecs lose task accuracy at low bitrates, so new codecs exploiting the redundancy structure of transformer features are needed.
  • Because feature MSE does not track task accuracy for textual features, semantic distortion metrics must be developed before LLM feature coding can be properly optimized.

Reading between the lines

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

  • Extension: the observed vertical redundancy in DINOv2 features suggests a codec that predicts along the token dimension, or a learned transform matched to patch-based features, could beat DCT-based packing; the paper identifies the property but does not test such a codec.
  • Extension: BPFP could plausibly be applied to embeddings and latents beyond the three models, but the paper's dataset does not show how the metric behaves on other large-model families, so that transfer remains untested.
  • Extension: a codec trained on 40th-block DINOv2 features may waste bits on early-block features, since the paper's distribution analysis shows layer-dependent ranges; a cross-layer generalizability test would extend the existing cross-task analysis.
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

3 major / 4 minor

Summary. The paper introduces a new research area, large model feature coding, and contributes a dataset (LaMoFC) of intermediate features extracted from three large models (DINOv2, Llama3, SD3) across five tasks (image classification, semantic segmentation, depth estimation, common sense reasoning, text-to-image synthesis). It proposes a new bitrate metric, Bits Per Feature Point (BPFP), and defines task-specific evaluation heads. Two baseline codecs are introduced by adapting VVC (VTM) and Hyperprior to feature inputs, and a benchmark is reported. The stated goal is to provide a standardized testbed and unified test conditions to enable fair comparisons in future feature coding research.

Significance. If the described dataset and test conditions were fully realized, this would be a valuable community resource. The public release of code and data is a concrete strength, as is the proposal of BPFP to decouple bitrate from pixel-based conventions. The dataset covers model types and modalities that are genuinely under-explored in the feature coding literature. However, the central claim of 'unified test conditions' is currently undermined by baseline-specific pre-processing, and the benchmark comparison is therefore not as clean as advertised. The dataset and BPFP contribution can stand after revision, but the benchmark conclusions need to be re-examined or re-framed.

major comments (3)
  1. [Sec. 6.1.1, Table 4] The paper claims in Sec. 5 that it establishes unified test conditions enabling fair comparisons, but Sec. 6.1.1 states that 'different truncations are used for the two baselines for their distinct coding strategies,' and Table 4 shows different truncation ranges for the same task (e.g., Cls: [-20,20] for VTM vs [-5,5] for Hyperprior; Seg: [-20,20] vs [-5,5]; SPDM4: [-20,20] vs [-10,10]). Since truncation is a lossy pre-processing step, the two baselines are coding different inputs at the same reported BPFP, so the head-to-head comparison in Table 3 is not a fair test of codec capability. Specifically, the conclusion that the VTM baseline outperforms the Hyperprior baseline may be an artifact of Hyperprior's more aggressive truncation rather than of the coding method itself. Fixing truncation ranges as part of the test conditions, or reporting results across a common set of truncation ranges, is necessary to support the 'unified' and 'fair comparison' claims.
  2. [Sec. 4.1.3] The source data selection is biased toward samples that are already correctly handled by the feature-extraction model. For Cls, the text says 'each representing a unique class accurately classified by DINOv2,' and for CSR, 'chosen for their longest input prompts and correct prediction by Llama3.' This guarantees a 100% oracle accuracy for these tasks, which inflates the initial accuracy and removes headroom for measuring accuracy degradation. It also may not reflect the full distribution of inputs encountered in real deployments, where misclassified or ambiguous samples are common. The authors should justify this choice explicitly or include a stratified sample with both correctly and incorrectly classified examples, and analyze the impact on the reported rate-accuracy curves.
  3. [Sec. 4.1 and Table 3] No error bars, confidence intervals, or repeated-run statistics are reported anywhere in the benchmark. Given the small sample sizes (80 for depth estimation, 100 for segmentation, 500 for the others), the differences between the two baselines at comparable bitrates (e.g., Cls at BPFP~2: 99.80 vs 93.20; Seg at BPFP~1.7: 80.42 vs 77.96) may be within sampling variability. Reporting standard deviations across multiple random seeds or bootstrap intervals would substantially strengthen the empirical claims and help readers assess whether the observed differences are significant.
minor comments (4)
  1. [Sec. 4.1.2] The notation SPDS, SPDM, SPG, and SPH is introduced in prose but Table 1 abbreviates these as split points without a cross-reference to the definitions; consider adding a dedicated column or a footnote in Table 1 to improve readability.
  2. [Sec. 4.2] Figures 2 and 3 are qualitative visualizations of distributions and DCT energy. To make the redundancy analysis more concrete, consider adding a quantitative measure such as the fraction of energy in the DC coefficient or a spatial autocorrelation statistic.
  3. [Sec. 6.2] The acronyms R-A and D-A are used in headings and figure captions but are not expanded in the main text; please define them at first use in Sec. 6.2 and Sec. 6.3.
  4. [Sec. 6.1.1] The sentence 'Different truncations are used for the two baselines for their distinct coding strategies' is an explicit acknowledgment of the fairness issue; even if this is a deliberate design choice, the paper should explain why this is acceptable under the 'unified test conditions' claim, rather than leaving the reader to reconcile the two statements.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the benchmark claims are supported by direct measurement with independently defined metrics and external codecs.

full rationale

The paper's central outputs are a dataset, a bitrate metric (BPFP = total bits divided by the number of feature points, Sec. 5.1), fixed task heads and accuracy metrics (Table 2), and baseline rate-accuracy measurements (Table 3). None of these quantities is defined in terms of the results they are used to produce: BPFP is a ratio of measured bit counts to a fixed feature-point count, task accuracy is evaluated with fixed task heads from the cited model papers, and the VTM and Hyperprior baselines are standard external codecs with stated configurations. The Hyperprior loss L = BPFP + lambda * ||Fo - Fr||^2 is a conventional rate-distortion objective whose lambda sweeps operating points; it is not fitted to make a particular outcome true. The paper does invoke several prior works by the same authors as examples of existing feature coding, but none is load-bearing for the dataset, metric, or benchmark conclusions; the central claims do not reduce to a self-citation chain. The notable methodological weakness is an internal inconsistency in the claimed unified test conditions: Sec. 6.1.1 says 'different truncations are used for the two baselines for their distinct coding strategies,' and Table 4 assigns different truncation ranges for the same task (e.g., Cls [-20,20] for VTM versus [-5,5] for Hyperprior). This is a validity confound for the cross-baseline comparison 'VTM baseline outperforms the Hyperprior baseline,' and it weakens the 'unified' claim, but it is not circularity: the comparison is still a measurement, not a result forced by definition or by fitted inputs. Accordingly, no circular step meets the evidence threshold.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

The benchmark depends on hand-chosen truncation ranges and quantization depth, and on the assumption that the selected models, tasks, and split points represent the space of large model features. These are reasonable for a first benchmark but are not derived from any external principle.

free parameters (2)
  • Truncation ranges per task/baseline = Cls/Seg: [-20,20]; Dpt: [-1,1] to [-20,20] by layer; CSR: [-5,5]; TTI: [-5.79,4.46]; Hyperprior uses tighter ranges…
    Hand-chosen in Table 4 to remove outliers from original feature ranges; they directly affect the bitrate-accuracy tradeoff of both baselines.
  • Quantization bit depth = 10 bits
    Uniform quantization to 10-bit integers is a design choice in Sec. 6.1.1; it sets the maximum fidelity of the pre-processing.
assumptions (3)
  • domain assumption The selected split points for DINOv2, Llama3, and SD3 yield features representative of large model distributed deployment.
    Invoked in Sec. 4.1.1 and Table 1; if other split points produce very different feature statistics, the benchmark's conclusions may not generalize.
  • domain assumption The fixed simple task heads and metrics in Table 2 accurately reflect downstream task performance.
    Sec. 5.2 fixes task heads to isolate codec effects, but a linear head may not capture the full sensitivity of real downstream models to feature distortion.
  • domain assumption Image codecs (VVC and Hyperprior) adapted via truncation and packing are reasonable baselines for large model feature coding.
    The paper uses this premise throughout Sec. 6; it is plausible but not derived from feature statistics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Feature Coding in the Era of Large Models: Dataset, Test Conditions, and Benchmark." pith.science (2026). https://pith.science/paper/UQ2SUXCF

@misc{pith2026241204307,
  author       = {Pith},
  title        = {Pith review of: Feature Coding in the Era of Large Models: Dataset, Test Conditions, and Benchmark},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UQ2SUXCF}},
  note         = {Machine review of arXiv:2412.04307}
}
read the original abstract

Large models have achieved remarkable performance across various tasks, yet they incur significant computational costs and privacy concerns during both training and inference. Distributed deployment has emerged as a potential solution, but it necessitates the exchange of intermediate information between model segments, with feature representations serving as crucial information carriers. To optimize information exchange, feature coding is required to reduce transmission and storage overhead. Despite its importance, feature coding for large models remains an under-explored area. In this paper, we draw attention to large model feature coding and make three fundamental contributions. First, we introduce a comprehensive dataset encompassing diverse features generated by three representative types of large models. Second, we establish unified test conditions, enabling standardized evaluation pipelines and fair comparisons across future feature coding studies. Third, we introduce two baseline methods derived from widely used image coding techniques and benchmark their performance on the proposed dataset. These contributions aim to provide a foundation for future research and inspire broader engagement in this field. To support a long-term study, all source code and the dataset are made available at \href{https://github.com/chansongoal/LaMoFC}{https://github.com/chansongoal/LaMoFC}.

Figures

Figures reproduced from arXiv: 2412.04307 by the authors.

Figure 1
Figure 1. Three application scenarios for feature coding in large model deployments. Cloud-centralized training: Features generated by foundation models are encoded and stored in a feature database. This database supports various downstream tasks, reducing computational demands by avoiding repeated inference of large foundation models. Cloud-edge distributed training: To protect user data privacy, raw data is first processed … view at source ↗
Figure 2
Figure 2. Frequency and CDF comparisons between features in the proposed dataset and commonly used existing features. The proposed [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Visualization of the original feature blocks and their corresponding DCT blocks (absolute values). In the feature domain, the [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The pipeline of the proposed baseline methods. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Rate-accuracy-drop (R-AD) comparisons among differ [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Compressed Feature Quality Assessment: Dataset and Baselines

    cs.CV 2025-06 conditional novelty 7.0 of 10

    The first compressed feature quality assessment benchmark is released, and three standard similarity metrics are shown to correlate inconsistently with task-level semantic distortion.

  2. Visual Token Codec: Unleashing Spatial Redundancy for ViT Feature Coding

    cs.CV 2026-08 conditional novelty 6.0 of 10

    The Visual Token Codec compresses ViT intermediate features by entropy-coding patch tokens on their native grid instead of a flattened sequence, reducing bitrate by 15.7x to 37.4x at 90% of uncompressed performance ov...

  3. DT-UFC: Universal Large Model Feature Coding via Peaky-to-Balanced Distribution Transformation

    cs.MM 2025-06 conditional novelty 4.0 of 10

    A per-model scalar quantization transform aligns heterogeneous feature distributions so that a single learned codec can compress features from LLaMA3, DINOv2, and Stable Diffusion 3 with better rate-accuracy than task...

  4. Cross-architecture universal feature coding via distribution alignment

    cs.CV 2025-06 conditional novelty 4.0 of 10

    A unified feature codec for CNN and ViT features, built with format and value alignment, beats an architecture-specific baseline on ImageNet classification.

Reference graph

Works this paper leans on

52 extracted references · 36 canonical work pages · cited by 4 Pith papers

  1. [1]

    Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston

    Johannes Ball ´e, David C. Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston. Variational image compression with a scale hyperprior. ArXiv, abs/1802.01436, 2018. 2

    arXiv 2018

  2. [2]

    Sullivan, and Jens-Rainer Ohm

    Benjamin Bross, Ye-Kui Wang, Yan Ye, Shan Liu, Jianle Chen, Gary J. Sullivan, and Jens-Rainer Ohm. Overview of the versatile video coding (VVC) standard and its applica- tions. IEEE Transactions on Circuits and Systems for Video Technology, 31(10):3736–3764, 2021. 2

    work page 2021

  3. [3]

    High efficient 3D convolution feature compression

    Yangang Cai, Peiyin Xing, and Xuesong Gao. High efficient 3D convolution feature compression. IEEE Transactions on Circuits and Systems for Video Technology, pages 1–1, 2022. 2

    work page 2022

  4. [4]

    When federated learning meets privacy- preserving computation

    Jingxue Chen, Hang Yan, Zhiyuan Liu, Min Zhang, Hu Xiong, and Shui Yu. When federated learning meets privacy- preserving computation. ACM Computing Surveys, 56(12): 1–36, 2024. 1, 3

    work page 2024

  5. [5]

    End-to-end learned scalable multilayer feature compression for machine vision tasks

    Qiaoxi Chen, Changsheng Gao, and Dong Liu. End-to-end learned scalable multilayer feature compression for machine vision tasks. In ICIP, pages 1781–1787, 2024. 2

    work page 2024

  6. [6]

    Toward intelligent sensing: Inter- mediate deep feature compression

    Zhuo Chen, Kui Fan, Shiqi Wang, Lingyu Duan, Weisi Lin, and Alex Chichung Kot. Toward intelligent sensing: Inter- mediate deep feature compression. IEEE Transactions on Image Processing, 29:2230–2243, 2020. 1

    work page 2020

  7. [7]

    Hyomin Choi and Ivan V . Baji´c. Latent-space scalability for multi-task collaborative intelligence. In ICIP, pages 3562– 3566, 2021. 2

    work page 2021

  8. [8]

    ImageNet: a large-scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. ImageNet: a large-scale hierarchical image database. In CVPR, pages 248–255, 2009. 5

    work page 2009

Show all 52 references
  1. [9]

    Upsampling DINOv2 features for unsupervised vi- sion tasks and weakly supervised materials segmentation

    Ronan Docherty, Antonis Vamvakeros, and Samuel J Cooper. Upsampling DINOv2 features for unsupervised vi- sion tasks and weakly supervised materials segmentation. arXiv preprint arXiv:2410.19836, 2024. 3

    2024 arXiv

  2. [10]

    Video coding for machines: A paradigm of collab- orative compression and intelligent analytics.IEEE Transac- tions on Image Processing, 29:8680–8695, 2020

    Lingyu Duan, Jiaying Liu, Wenhan Yang, Tiejun Huang, and Wen Gao. Video coding for machines: A paradigm of collab- orative compression and intelligent analytics.IEEE Transac- tions on Image Processing, 29:8680–8695, 2020. 2

    2020

  3. [11]

    The Llama 3 herd of models

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Ab- hishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The Llama 3 herd of models. arXiv preprint arXiv:2407.21783,

    arXiv

  4. [12]

    Scaling recti- fied flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas M ¨uller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling recti- fied flow transformers for high-resolution image synthesis. In ICML, 2024. 2, 4, 6

    2024

  5. [13]

    Image coding for machines with omnipotent feature learn- ing

    Ruoyu Feng, Xin Jin, Zongyu Guo, Runsen Feng, Yixin Gao, Tianyu He, Zhizheng Zhang, Simeng Sun, and Zhibo Chen. Image coding for machines with omnipotent feature learn- ing. In ECCV, pages 510–528. Springer, 2022. 2

    2022

  6. [14]

    LLM-based edge intelligence: A comprehensive survey on architectures, applications, security and trustworthiness

    Othmane Friha, Mohamed Amine Ferrag, Burak Kantarci, Burak Cakmak, Arda Ozgun, and Nassira Ghoualmi-Zine. LLM-based edge intelligence: A comprehensive survey on architectures, applications, security and trustworthiness. IEEE Open Journal of the Communications Society, 5:5799– ...

    2024

  7. [15]

    Towards task-generic image compression: A study of semantics- oriented metrics

    Changsheng Gao, Dong Liu, Li Li, and Feng Wu. Towards task-generic image compression: A study of semantics- oriented metrics. IEEE Transactions on Multimedia, 25:721– 735, 2023. 2

    2023

  8. [16]

    DMOFC: discrimination metric-optimized feature com- pression

    Changsheng Gao, Yiheng Jiang, Li Li, Dong Liu, and Feng Wu. DMOFC: discrimination metric-optimized feature com- pression. In PCS, pages 1–5, 2024. 1

    2024

  9. [17]

    Rethinking the joint optimization in video coding for machines: A case study

    Changsheng Gao, Zhuoyuan Li, Li Li, Dong Liu, and Feng Wu. Rethinking the joint optimization in video coding for machines: A case study. In DCC, pages 556–556, 2024. 2

    2024

  10. [18]

    IMOFC: identity-level metric optimized fea- ture compression for identification tasks

    Changsheng Gao, Yiheng Jiang, Siqi Wu, Yifan Ma, Li Li, and Dong Liu. IMOFC: identity-level metric optimized fea- ture compression for identification tasks. IEEE Transactions on Circuits and Systems for Video Technology, 35(2):1855– 1869, 2025. 1

    2025

  11. [19]

    Chal- lenges and applications of large language models

    Jean Kaddour, Joshua Harris, Maximilian Mozes, Herbie Bradley, Roberta Raileanu, and Robert McHardy. Chal- lenges and applications of large language models. arXiv preprint arXiv:2307.10169, 2023. 1

    2023 arXiv

  12. [20]

    Bridging compressed image latents and multimodal large language models

    Chia-Hao Kao, Cheng Chien, Yu-Jen Tseng, Yi-Hsin Chen, Alessandro Gnutti, Shao-Yuan Lo, Wen-Hsiao Peng, and Riccardo Leonardi. Bridging compressed image latents and multimodal large language models. arXiv preprint arXiv:2407.19651, 2024. 2

    2024 arXiv

  13. [21]

    End-to-end learnable multi-scale feature compression for VCM

    Yeongwoong Kim, Hyewon Jeong, Janghyun Yu, Younhee Kim, Jooyoung Lee, Se Yoon Jeong, and Hui Yong Kim. End-to-end learnable multi-scale feature compression for VCM. IEEE Transactions on Circuits and Systems for Video Technology, pages 1–1, 2023. 2

    2023

  14. [22]

    From big to small: Multi-scale local planar guidance for monocular depth estimation

    Jin Han Lee, Myung-Kyu Han, Dong Wook Ko, and Il Hong Suh. From big to small: Multi-scale local planar guidance for monocular depth estimation. arXiv preprint arXiv:1907.10326, 2019. 5

    1907 arXiv

  15. [23]

    GShard: Scaling giant models with conditional computation and automatic sharding

    Dmitry Lepikhin, HyoukJoong Lee, Yuanzhong Xu, Dehao Chen, Orhan Firat, Yanping Huang, Maxim Krikun, Noam Shazeer, and Zhifeng Chen. GShard: Scaling giant models with conditional computation and automatic sharding. arXiv preprint arXiv:2006.16668, 2020. 1

    2006 arXiv

  16. [24]

    Attention-based variable-size feature compression module for edge inference

    Shibao Li, Chenxu Ma, Yunwu Zhang, Longfei Li, Chengzhi Wang, Xuerong Cui, and Jianhang Liu. Attention-based variable-size feature compression module for edge inference. The Journal of Supercomputing, 2023. 2

    2023

  17. [25]

    Ustc-td: A test dataset and bench- mark for image and video coding in 2020s

    Zhuoyuan Li, Junqi Liao, Chuanbo Tang, Haotian Zhang, Yuqi Li, Yifan Bian, Xihua Sheng, Xinmin Feng, Yao Li, Changsheng Gao, et al. Ustc-td: A test dataset and bench- mark for image and video coding in 2020s. arXiv preprint arXiv:2409.08481, 2024. 2

    2024 arXiv

  18. [26]

    Object segmentation-assisted inter prediction for versatile video coding

    Zhuoyuan Li, Zikun Yuan, Li Li, Dong Liu, Xiaohu Tang, and Feng Wu. Object segmentation-assisted inter prediction for versatile video coding. IEEE Transactions on Broadcast- ing, 2024. 2

    2024

  19. [27]

    Learnt mutual feature compression for machine vision

    Tie Liu, Mai Xu, Shengxi Li, Chaoran Chen, Li Yang, and Zhuoyi Lv. Learnt mutual feature compression for machine vision. In ICASSP, pages 1–5, 2023. 2

    2023

  20. [28]

    Preprocessing enhanced image compression for ma- chine vision

    Guo Lu, Xingtong Ge, Tianxiong Zhong, Qiang Hu, and Jing Geng. Preprocessing enhanced image compression for ma- chine vision. IEEE Transactions on Circuits and Systems for Video Technology, pages 1–1, 2024. 2

    2024

  21. [29]

    Lingjuan Lyu, Han Yu, Xingjun Ma, Chen Chen, Lichao Sun, Jun Zhao, Qiang Yang, and Philip S. Yu. Privacy and ro- bustness in federated learning: Attacks and defenses. IEEE Transactions on Neural Networks and Learning Systems, 35 (7):8726–8746, 2024. 3

    2024

  22. [30]

    Feature compression with 3d sparse con- volution

    Yifan Ma, Changsheng Gao, Qiaoxi Chen, Li Li, Dong Liu, and Xiaoyan Sun. Feature compression with 3d sparse con- volution. In VCIP, pages 1–5, 2024. 2

    2024

  23. [31]

    Perceptual image compression with con- ditional diffusion transformers

    Rui Mao, Xinmin Feng, Changsheng Gao, Li Li, Dong Liu, and Xiaoyan Sun. Perceptual image compression with con- ditional diffusion transformers. In VCIP, pages 1–5, 2024. 2

    2024

  24. [32]

    Video feature compression for machine tasks

    Kiran Misra, Tianying Ji, Andrew Segall, and Frank Bossen. Video feature compression for machine tasks. In ICME, pages 1–6, 2022. 2

    2022

  25. [33]

    DINOv2: Learning robust visual features without supervi- sion

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. DINOv2: Learning robust visual features without supervi- sion. arXiv preprint arXiv:2304.07193, 2023. 2, 4, 6

    2023 arXiv

  26. [34]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In ICML, pages 8748–8763. PMLR, 2021. 3

    2021

  27. [35]

    Vnvc: A versatile neural video coding framework for efficient human- machine vision

    Xihua Sheng, Li Li, Dong Liu, and Houqiang Li. Vnvc: A versatile neural video coding framework for efficient human- machine vision. IEEE Transactions on Pattern Analysis and Machine Intelligence, 46(7):4579–4596, 2024. 2

    2024

  28. [36]

    Eva-CLIP: improved training techniques for clip at scale

    Quan Sun, Yuxin Fang, Ledell Wu, Xinlong Wang, and Yue Cao. Eva-CLIP: improved training techniques for clip at scale. arXiv preprint arXiv:2303.15389, 2023. 3

    2023 arXiv

  29. [37]

    Deep feature compression using spatio-temporal arrangement toward col- laborative intelligent world

    Satoshi Suzuki, Shoichiro Takeda, Motohiro Takagi, Ryuichi Tanida, Hideaki Kimata, and Hayaru Shouno. Deep feature compression using spatio-temporal arrangement toward col- laborative intelligent world. IEEE Transactions on Circuits and Systems for Video Technology, 32(6):3934...

    2022

  30. [38]

    FedBERT: when federated learning meets pre-training

    Yuanyishu Tian, Yao Wan, Lingjuan Lyu, Dezhong Yao, Hai Jin, and Lichao Sun. FedBERT: when federated learning meets pre-training. ACM Transactions on Intelligent Systems and Technology, 13(4):1–26, 2022. 1

    2022

  31. [39]

    Non- semantics suppressed mask learning for unsupervised video semantic compression

    Yuan Tian, Guo Lu, Guangtao Zhai, and Zhiyong Gao. Non- semantics suppressed mask learning for unsupervised video semantic compression. In ICCV, pages 13564–13576, 2023. 2

    2023

  32. [40]

    SMC++: masked learning of unsupervised video semantic compression

    Yuan Tian, Guo Lu, and Guangtao Zhai. SMC++: masked learning of unsupervised video semantic compression. arXiv preprint arXiv:2406.04765, 2024. 2

    2024

  33. [41]

    Free-VSC: free semantics from visual foundation models for unsupervised video semantic compression

    Yuan Tian, Guo Lu, and Guangtao Zhai. Free-VSC: free semantics from visual foundation models for unsupervised video semantic compression. In ECCV, pages 163–183,

  34. [42]

    Llama: Open and efficient foundation language models

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timoth´ee Lacroix, Baptiste Rozi`ere, Naman Goyal, Eric Hambro, Faisal Azhar, et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971, 2023. 2

    2023 arXiv

  35. [43]

    Split learning for health: Distributed deep learning without sharing raw patient data

    Praneeth Vepakomma, Otkrist Gupta, Tristan Swedish, and Ramesh Raskar. Split learning for health: Distributed deep learning without sharing raw patient data. arXiv preprint arXiv:1812.00564, 2018. 1, 3

    2018 arXiv

  36. [44]

    Towards analysis- friendly face representation with scalable feature and texture compression

    Shurun Wang, Shiqi Wang, Wenhan Yang, Xinfeng Zhang, Shanshe Wang, Siwei Ma, and Wen Gao. Towards analysis- friendly face representation with scalable feature and texture compression. IEEE Transactions on Multimedia , 24:3169– 3181, 2022. 1

    2022

  37. [45]

    NExT-GPT: Any-to-any multimodal LLM

    Shengqiong Wu, Hao Fei, Leigang Qu, Wei Ji, and Tat-Seng Chua. NExT-GPT: Any-to-any multimodal LLM. In ICML, pages 53366–53397, 2024. 5

    2024

  38. [46]

    On protecting the data privacy of large language models (LLMs): A survey

    Biwei Yan, Kun Li, Minghui Xu, Yueyan Dong, Yue Zhang, Zhaochun Ren, and Xiuzhen Cheng. On protecting the data privacy of large language models (LLMs): A survey. arXiv preprint arXiv:2403.05156, 2024. 3

    2024 arXiv

  39. [47]

    SSSIC: Semantics-to-signal scalable image coding with learned structural representations

    Ning Yan, Changsheng Gao, Dong Liu, Houqiang Li, Li Li, and Feng Wu. SSSIC: Semantics-to-signal scalable image coding with learned structural representations. IEEE Trans- actions on Image Processing, 30:8939–8954, 2021. 2

    2021

  40. [48]

    Video coding for machines: Compact vi- sual representation compression for intelligent collaborative analytics

    Wenhan Yang, Haofeng Huang, Yueyu Hu, Ling-Yu Duan, and Jiaying Liu. Video coding for machines: Compact vi- sual representation compression for intelligent collaborative analytics. IEEE Transactions on Pattern Analysis and Ma- chine Intelligence, pages 1–18, 2024. 2

    2024

  41. [49]

    Open- FedLLM: Training large language models on decentralized private data via federated learning

    Rui Ye, Wenhao Wang, Jingyi Chai, Dihan Li, Zexi Li, Yinda Xu, Yaxin Du, Yanfeng Wang, and Siheng Chen. Open- FedLLM: Training large language models on decentralized private data via federated learning. In ACM SIGKDD, pages 6137–6147, 2024. 1, 3

    2024

  42. [50]

    All-in-one image coding for joint human-machine vision with multi- path aggregation

    Xu Zhang, Peiyao Guo, Ming Lu, and Zhan Ma. All-in-one image coding for joint human-machine vision with multi- path aggregation. arXiv preprint arXiv:2409.19660 , 2024. 2

    2024 arXiv

  43. [51]

    MSFC: Deep feature compression in multi-task network

    Zhicong Zhang, Mengyang Wang, Mengyao Ma, Jiahui Li, and Xiaopeng Fan. MSFC: Deep feature compression in multi-task network. In ICME, pages 1–6, 2021. 2

    2021

  44. [52]

    Safely learning with private data: A federated learning framework for large language model

    Jiaying Zheng, Hainan Zhang, Lingxiang Wang, Wangjie Qiu, Hongwei Zheng, and Zhiming Zheng. Safely learning with private data: A federated learning framework for large language model. arXiv preprint arXiv:2406.14898, 2024. 1, 3

    2024 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.