Papers › Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling

Designing BERT for Convolutional Networks: Sparse and Hierarchical Masked Modeling

9 Jan 2023arXiv:2301.03580archive 2025-07-28

Keyu Tian, Yi Jiang, Qishuai Diao, Chen Lin, LiWei Wang, Zehuan Yuan

We identify and overcome two key obstacles in extending the success of BERT-style pre-training, or the masked image modeling, to convolutional networks (convnets): (i) convolution operation cannot handle irregular, random-masked input images; (ii) the single-scale nature of BERT pre-training is inconsistent with convnet's hierarchical structure. For (i), we treat unmasked pixels as sparse voxels of 3D point clouds and use sparse convolution to encode. This is the first use of sparse convolution for 2D masked modeling. For (ii), we develop a hierarchical decoder to reconstruct images from multi-scale encoded features. Our method called Sparse masKed modeling (SparK) is general: it can be used directly on any convolutional model without backbone modifications. We validate it on both classical (ResNet) and modern (ConvNeXt) models: on three downstream tasks, it surpasses both state-of-the-art contrastive learning and transformer-based masked modeling by similarly large margins (around +1.0%). Improvements on object detection and instance segmentation are more substantial (up to +3.5%), verifying the strong transferability of features learned. We also find its favorable scaling behavior by observing more gains on larger models. All this evidence reveals a promising future of generative pre-training on convnets. Codes and models are released at https://github.com/keyu-tian/SparK.

PaperPDFCodeCode Syntology ran

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 14 code samples harvested from 2 repositories linked to this paper: 13 ran (of which 0 constructed an object rather than computing a result; 11 with no instrument failure: 0 honoured, 0 violated, 11 with no contract checked; 2 where Syntology's instrument failed) · 1 unverified. Where Syntology's instrument failed: 2 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: official repository: 10 samples from 1 repository, 9 ran (7 with no instrument failure; 2 where Syntology's instrument failed); community (archive-listed): 4 samples from 1 repository, 4 ran (4 with no instrument failure; 0 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written between 2026-09-22 and 2026-09-25 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

keyu-tian/spark officialmentioned in papermentioned on GitHubpytorchMITSyntology: 10 samples, 9 ran (2 instrument-failed) · 1 unverified report
Factral/UnMix-NeRF mentioned on GitHubjaxApache-2.0Syntology: 4 samples, 4 ran (0 instrument-failed) · 0 unverified report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

14 samples harvested: 13 ran (of which 0 constructed an object rather than computing a result; 11 with no instrument failure: 0 honoured, 0 violated, 11 with no contract checked; 2 where Syntology's instrument failed) · 1 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

2ran · our draft was wrong
11ran
1unverified

Licence: 2 of the 14 samples are pointer only, meaning Syntology does not serve that copy's text. For 2 of them, this paper's copy is under a permissive licence and cleared, but the stored copy of the same code that Syntology serves is not; the licence label's tooltip gives Syntology's recorded reason. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from 2 repositories linked to this paper, official or community; the list below is grouped by repository, each group headed by its name and what it is to the paper. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity. A census tag appears when Syntology's census ran implementations with that name on one shared input per bucket: it gives how many were compared, from how many papers, and how many distinct outputs they produced, and names this sample's bucket when this sample was one of them.

keyu-tian/spark official repository 10 samples, 9 ran (2 instrument-failed) · 1 unverified
get_num_layer_layer_wise in 5 papers downstream_mmdet/mmcv_custom/layer_decay_optimizer_constructor.py ran fingerprinted MIT (permissive) · aa8d49041d9f8442 · report fixture: signature var_name:str/0/str, num_max_layer:int/0/int · returned int · level recorded 2026-09-24
get_num_layer_stage_wise in 5 papers downstream_mmdet/mmcv_custom/layer_decay_optimizer_constructor.py ran fingerprinted MIT (permissive) · be134d6ef63b5192 · report fixture: signature var_name:str/0/str, num_max_layer:int/0/int · returned int · level recorded 2026-09-24
get_param_groups in 12 papers downstream_imagenet/lr_decay.py ran MIT (permissive) · ad373c0953d230ed · report fixture: signature model:callable/0/?, nowd_keys:tuple/1/str, lr_scale:float/0/float · returned list · level recorded 2026-09-25
load_checkpoint in 277 papers downstream_imagenet/util.py ran MIT (permissive) · 7475c50600ea1392 · report fixture: signature resume_from:str/0/str, model_without_ddp:callable/0/?, ema_module:callable/0/?, optimizer:SGD/0/? · returned tuple · level recorded 2026-09-25
lr_factor_func downstream_d2/lr_decay.py ran MIT (permissive) · 800af14f9c740324 · report fixture: signature para_name:str/0/str, is_resnet50:bool/0/bool, dec:float/0/float, debug:bool/0/bool · returned float · level recorded 2026-09-25
lr_wd_annealing downstream_imagenet/lr_decay.py ran MIT (permissive) · f921a5666329b4eb · report fixture: signature optimizer:SGD/0/?, peak_lr:float/0/float, wd:float/0/float, cur_it:int/0/int, wp_it:float/0/float, max_it:int/0/int · returned tuple · level recorded 2026-09-25
mixup_target in 22 papers downstream_imagenet/mixup.py ran MIT (permissive) here; the stored copy is pointer only · 8733e0d87f232d11 · report fixture: signature target:Tensor/1/int, num_classes:int/0/int, lam:float/0/float, smoothing:float/0/float, device:str/0/str · returned Tensor · level recorded 2026-09-24
one_hot in 150 papers
census: 18 implementations from 25 papers, 10 distinct outputs across 6 buckets, one shared input per bucket this sample: bucket 1 of 6, arg 1: rank 1, kind int, dtype int64
downstream_imagenet/mixup.py ran · our draft was wrong MIT (permissive) · 4728dc2025227bba · report fixture: signature x:Tensor/1/int, num_classes:int/0/int, on_value:float/0/float, off_value:float/0/float, device:str/0/str · returned Tensor · level recorded 2026-09-22
rand_bbox in 63 papers downstream_imagenet/mixup.py ran · our draft was wrong MIT (permissive) here; the stored copy is pointer only · 3913c620f6926f42 · report fixture: signature img_shape:tuple/1/int, lam:float/0/float, margin:float/0/float, count:NoneType/0/? · returned tuple · level recorded 2026-09-24
create_classification_dataset downstream_imagenet/data.py unverified MIT (permissive) · 0e865e01e71cf326 · report
Factral/UnMix-NeRF community (archive-listed) 4 samples, 4 ran (0 instrument-failed) · 0 unverified
component_x unmixnerf/utils/spec_to_rgb.py ran fingerprinted Apache-2.0 (permissive) · a883944500f3f33d · report fixture: signature x:ndarray/1/float · returned ndarray · level recorded 2026-09-25
component_y unmixnerf/utils/spec_to_rgb.py ran fingerprinted Apache-2.0 (permissive) · 8da57cf56b2fdcf1 · report fixture: signature x:ndarray/1/float · returned ndarray · level recorded 2026-09-25
g in 21 papers
census: 6 implementations from 6 papers, 6 distinct outputs across 3 buckets, one shared input per bucket
unmixnerf/utils/spec_to_rgb.py ran fingerprinted Apache-2.0 (permissive) · 558f6d7e47654006 · report fixture: signature x:ndarray/1/float, alpha:float/0/float, mu:float/0/float, sigma1:float/0/float, sigma2:float/0/float · returned ndarray · level recorded 2026-09-25
get_weights_spectral unmixnerf/unmix_renderer.py ran fingerprinted Apache-2.0 (permissive) · 36ae228f14629970 · report fixture: signature deltas:Tensor/3/float, densities:Tensor/3/float · returned Tensor · level recorded 2026-09-25

Tasks

2D Object DetectionContrastive LearningDecoderImage ClassificationInstance SegmentationObject DetectionSelf-Supervised Image ClassificationSemantic Segmentationobject-detection

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Image Classification ImageNet SparK (ConvNeXt-Large, 384) Number of params 198M #183 of 1060 Archive leaderboard report
Image Classification ImageNet SparK (ConvNeXt-Large, 384) Top 1 Accuracy 86.0% #183 of 1060 Archive leaderboard report
Instance Segmentation COCO 2017 val SparK (ConvNeXt V1-B Mask R-CNN) AP 45.1 #1 of 4 Archive leaderboard report
Instance Segmentation COCO 2017 val SparK (ConvNeXt V1-B Mask R-CNN) mask AP 45.1 #1 of 4 Archive leaderboard report
Instance Segmentation COCO 2017 val SparK (ConvNeXt V1-B Mask R-CNN) mask AP* 45.1 #1 of 4 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) SparK (ConvNeXt-Large, 384) Number of Params 198M #18 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) SparK (ConvNeXt-Large, 384) Top 1 Accuracy 86.0% #18 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) SparK (ConvNeXt-Large) Number of Params 198M #23 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) SparK (ConvNeXt-Large) Top 1 Accuracy 85.4% #23 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ConvNeXt-Base (SparK pre-training) Number of Params 89M #28 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ConvNeXt-Base (SparK pre-training) Top 1 Accuracy 84.8% #28 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ConvNeXt-Small (SparK pre-training) Number of Params 50M #37 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ConvNeXt-Small (SparK pre-training) Top 1 Accuracy 84.1% #37 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-200 (SparK pre-training) Number of Params 65M #47 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-200 (SparK pre-training) Top 1 Accuracy 83.1% #47 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-152 (SparK pre-training) Number of Params 60M #49 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-152 (SparK pre-training) Top 1 Accuracy 82.7% #49 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-101 (SparK pre-training) Number of Params 44M #52 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-101 (SparK pre-training) Top 1 Accuracy 82.2% #52 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-50 (SparK pre-training) Number of Params 26M #55 of 65 Archive leaderboard report
Self-Supervised Image Classification ImageNet (finetuned) ResNet-50 (SparK pre-training) Top 1 Accuracy 80.6% #55 of 65 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

AdamAttentionAttention DropoutBERTContrastive LearningConvolutionDense ConnectionsDropoutLayer NormalizationLinear LayerLinear Warmup With Linear DecayMulti-Head AttentionResidual ConnectionSoftmaxWeight DecayWordPiece

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections