Pith. sign in

REVIEW 6 major objections 5 minor 31 references

Pieceformer: Similarity-Driven Knowledge Transfer via Scalable Graph Transformer in VLSI

T0 review · 6 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Pieceformer claims that unlabeled VLSI circuits can be ranked by structural similarity with a partitioned linear-attention transformer, and that the resulting similarity score can JumpStart downstream EDA tasks, cutting partitioning…

desk verdict Worth a referee's time, but the core 24.9% claim needs a documented train/test split and a partitioned GIN baseline before it means anything. read the letter →

arxiv 2506.15907 v1 pith:XXXPGJTU submitted 2025-06-18 cs.LG cs.SYeess.SY

Hang Yang , Yusheng Hu , Yong Liu , Cong (Callie) Hao This is my paper · ORCID
classification cs.LGcs.SYeess.SY
keywords graphsimilarityVLSIdesignreuseself-supervisedlearningtransformerpartitionedtraininglinearattentionknowledgetransfercircuitnetlist
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

The paper claims that a large chip-design (VLSI) circuit can be judged similar to other circuits by an embedding learned entirely from graph structure, with no human labels, and that this similarity score can transfer a known good design solution to a new circuit. Pieceformer combines local message passing with a partitioned graph transformer built on linear attention, training on roughly 500-node subgraphs of real netlists and stitching the subgraph embeddings back together. On synthetic gate-level graphs this lowers mean absolute error in similarity ranking by 24.9% on average against the message-passing baseline, and on the real-world CircuitNet set it is the only tested method whose learned clusters separate every design group. In a Kernighan-Lin partitioning case study, using the most similar design's configuration as the starting point cuts runtime by 53% on 100-node graphs and up to 89% on 1000-node graphs. The importance, if the claim holds, is that design reuse no longer depends on expensive edit-distance computations or subjective similarity judgments.

What carries the argument

The load-bearing object is the hybrid MP+PGT encoder: a message-passing layer that injects local context, followed by a partitioned graph transformer with linear attention that attends within roughly 500-node subgraphs instead of the full graph. The partitioner splits each netlist in preprocessing, and the pipeline concatenates the per-subgraph embeddings to reconstruct a graph-level embedding; a contrastive loss maximizes mutual information between node-level and graph-level embeddings of the same graph. This division of labor removes the quadratic attention cost and memory blow-up that makes full transformers infeasible above about 1.1k nodes, and it also avoids the softmax saturation that degrades attention when too many nodes compete, which is why a partition size around 500 minimizes ranking error.

What would settle it

Generate pairs of synthetic VLSI graphs that are identical within every roughly 500-node partition but differ only in a few long-range edges that cross the chosen partition boundaries. If Pieceformer ranks these pairs as more similar than pairs differing by the same number of within-partition edits, then cross-partition structure is invisible to the learned similarity; if ranking MAE stays at the same level as for full-graph edits, the concatenated embedding preserves enough global structure.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that the bottleneck for similarity-driven design reuse is not the similarity definition or the data, but the encoder's receptive field: message-passing alone is too local, full graph transformers are too memory-hungry and suffer attention dilution on multi-thousand-node circuits, while a pipeline that partitions each graph into roughly 500-node subgraphs, runs a linear-attention transformer on each partition, prepends a message-passing layer, and concatenates the subgraph embeddings into one graph embedding under a contrastive self-supervised objective gives accurate rankings and clean design-group clusters. This encoder produces graph-level embeddings whose L2 distance orders synthetic graphs by edit-distance-like structural change and separates real 28nm design groups that other encoders mix up. The same embedding, used as a JumpStart source selector, accelerates KL partitioning by reusing the winning partition of the nearest design.

Load-bearing premise

The load-bearing premise is that chopping each circuit into roughly 500-node subgraphs and concatenating their embeddings keeps the global structure that determines similarity; because attention never crosses subgraph boundaries, any design similarity carried by cross-partition wiring is invisible to the model.

Editorial extensions

If this is right

  • Design libraries can be searched by structural similarity without any labeled similarity scores, so a new netlist can inherit partitioning, synthesis, or floorplan configurations from the nearest prior design.
  • Graph size stops being a hard barrier to transformer training on circuits: the partitioned pipeline handles netlists of tens of thousands of nodes on a single GPU while using more of the GPU's compute.
  • Ranking accuracy holds across a wide range of sizes, with a reported 40.7% MAE improvement over the message-passing baseline at 1,000 nodes and a 24.9% average reduction across all tested scales.
  • The KL partitioning case study indicates that the speedup grows with graph size, 53% at 100 nodes and up to 89% at 1,000 nodes, concentrating the benefit exactly where EDA iteration is most expensive.
  • Because the method is label-free and task-agnostic, the same pretrained similarity score can JumpStart any iterative EDA task with a parameterized initial state, including synthesis configuration reuse and power grid tuning.

Reading between the lines

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

  • A natural extension the paper leaves open is that, because attention never crosses partition boundaries, cross-partition wiring can only be encoded through the concatenated embedding; adding a lightweight global summary over partitions might capture long-range structure the current pipeline would miss.
  • Not tested in the paper: the 89% KL speedup is measured against a random initialization, so extrapolating to commercial EDA flows would require comparison against expert-tuned starts, a comparison the paper itself notes is complicated by a chicken-or-egg problem.
  • If the roughly 500-node partition sweet spot generalizes, the partitioned linear-attention recipe could transfer to other large-graph similarity tasks outside VLSI, such as program or social-network comparison, where labeled similarity is scarce.
  • A direct test of the partition-blindness assumption: generate synthetic edit operations that only rewire edges across the chosen partition boundaries, then check whether ranking MAE deteriorates; if it does, part of the similarity signal is an artifact of partition choice.
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

6 major / 5 minor

Summary. This paper introduces Pieceformer, a self-supervised graph similarity framework for VLSI design reuse, which combines a message-passing GNN with a partitioned linear transformer (PGT) and trains embeddings via the InfoGraph contrastive objective. Input graphs are split into METIS subgraphs of configurable size, and subgraph embeddings are concatenated to form the graph embedding. The authors evaluate on a synthetic dataset of base graphs with 15 derived graphs per base, reporting a 24.9% average MAE reduction over a GIN baseline in similarity ranking, and on the CircuitNet dataset, where they claim the only method to correctly separate all design groups in UMAP projections. A case study applies the similarity ranking to initialize a Kernighan-Lin partitioner, reporting up to 89% runtime reduction.

Significance. The problem of scalable, label-free graph similarity for VLSI is timely, and the partitioned training pipeline is a pragmatic contribution to transformer scalability. The self-supervised setting and the JumpStart formulation are potentially useful. However, the current experimental protocol does not establish that the reported gains reflect generalization: the synthetic evaluation lacks a described train/test split, the baseline is not matched in partitioning, and the CircuitNet clustering claim is qualitative. If the experiments are properly reconducted, the framework could be a solid contribution; as written, the central quantitative claims are not yet supported.

major comments (6)
  1. [V-B, Fig. 2] The synthetic evaluation does not specify which graphs are used for training versus ranking. Section V-A1 describes 16-graph groups (base G_B and 15 derived graphs), and Section V-B reports MAE after training, but the paper never states whether the InfoGraph encoder was trained on the same graphs that are later embedded and ranked. If training and evaluation share graph instances, the 24.9% MAE improvement over GIN may reflect per-graph memorization rather than a generalizable similarity function. Please provide an explicit train/test split (e.g., hold out entire graph groups, or reserve a subset of derived graphs), and report MAE on the held-out graphs. Without this, the headline result is not scientifically interpretable.
  2. [V-B, Fig. 2] The baseline GIN is not partitioned, so the reported improvement cannot be attributed to the hybrid MP+PGT architecture versus the partitioning scheme. To isolate the effect of partitioning, include a partitioned GIN baseline (GIN applied to the same METIS subgraphs with concatenated embeddings, or GIN+PGT without the transformer). Also report standard deviations across multiple random seeds or graph groups; Fig. 2 shows a single MAE value per scale with no error bars, so the 24.9% average reduction has no measure of variance.
  3. [V-B, Fig. 3] The partition size (500) is selected using the same synthetic evaluation data on which Fig. 2 reports MAE. Because the hyperparameter is tuned on the test set, the reported gains are optimistic. Please either use a separate validation split for partition-size selection or report MAE across a range of partition sizes with error bars, and state clearly which numbers are selected post hoc.
  4. [V-C, Fig. 4] The claim that Pieceformer is the only method to correctly cluster all design groups is based on visual inspection of UMAP projections. This is not a quantitative evaluation. Please report cluster-quality metrics computed on the graph embeddings (e.g., silhouette score, adjusted Rand index against the known design groups, or k-nearest-neighbor accuracy), and specify the UMAP hyperparameters (n_neighbors, min_dist) and any normalization steps. In addition, clarify whether the model used for CircuitNet embeddings was trained on the synthetic dataset or on CircuitNet; if the latter, describe the training set and ensure the clustering evaluation is not conducted on the training data.
  5. [V-D] The KL partitioning case study lacks essential details. First, it is not described how the partition of the most similar design is transferred to the target graph, which has a different node set; a mapping between nodes must be specified (e.g., by node degree or embedding alignment). Second, runtimes are reported as single numbers (53% and 89% reduction); KL is a randomized algorithm, so the comparison should report mean and standard deviation over many random initializations for both the JumpStart and random-start baselines. Third, the target graphs in this study have 100 and 1,000 nodes, which is far smaller than the 49k-node scale claimed in Table I, so the practical scalability claim is not demonstrated by this case study.
  6. [IV-C] Because attention is applied only within each METIS subgraph, the 'global attention' of PGT is not global across the graph; cross-partition dependencies are invisible to the encoder. The paper does not provide evidence that VLSI graph similarity is dominated by intra-partition structure. Please add an ablation or analysis (e.g., compare METIS partitioning with random partitioning, or add a cross-partition interaction layer) to justify this architectural choice, or revise the terminology to avoid overstating the model's global reach.
minor comments (5)
  1. [IV-B] The reference [5] cited for Performer is 'Masked language modeling for proteins via linearly scalable long-context transformers' (Choromanski et al., 2020), which is not the Performer paper. Please cite the correct source (e.g., Choromanski et al., 'Rethinking Attention with Performers', ICLR 2021).
  2. [V-A1] The amount of 'limited number of nodes/edges' in the edit operations is never quantified. Please specify the distribution or range of edit counts used to generate the derived graphs.
  3. [V-A1 and Conclusion] The abstract and contribution list claim the method is 'effective even with minimal training data (as few as four graphs in practice)', but no experiment with only four training graphs is reported. Either add such an experiment or remove the claim.
  4. [V-C and Table I] The text says CircuitNet contains over 20,000 samples, but Table I lists 50 graphs. Clarify whether the 50 graphs in Table I are a subset of the 20,000 samples or the entire dataset used in this paper.
  5. [Fig. 4] The runtime analysis is presented as a small table embedded in the figure, and the text references 'ΔRuntime' without defining the baseline for the percentage change. Please move this to a separate table with a clear baseline and annotate the units (s/epoch).

Circularity Check

1 steps flagged · score 4.0 of 10

Reported 24.9% MAE gain is partly circular: partition size is selected by minimizing MAE on the same synthetic set used to report the headline improvement.

  1. fitted input called prediction [Section V-B, Fig. 3 (partition-size selection) and Fig. 2 / Abstract (reported MAE)]
    "To evaluate the impact of partition size on model performance, we conduct more experiments on the synthetic dataset. As shown in Fig. 3, MAE is lowest when each partition contains around 500 nodes. Larger partitions degrade performance due to softmax saturation, where attention becomes “diluted” with too many nodes, similar to full graph training. All other evaluations use 500 as partition size."

    The partition size is a model configuration chosen by minimizing MAE on the synthetic evaluation set (Fig. 3). The same synthetic set and the same MAE metric are then used to report the model's headline result (Fig. 2: 'an average improvement of 24.9% over the baseline'; abstract: 'reduces mean absolute error (MAE) by 24.9%'). Because no train/validation/test split is described in Section V, the reported ranking accuracy is the same quantity that was optimized when selecting the partition size. The 'prediction' therefore partly reduces to the selection criterion, rather than being an independent out-of-sample test of the proposed model.

full rationale

The paper's derivation of the method (hybrid MP+PGT encoder trained with InfoGraph-style contrastive loss, METIS partitioning, linear transformer) is not circular: the training objective does not use the synthetic edit-count rankings or the CircuitNet group labels, and no load-bearing result is imported from the authors' prior work. The only concrete reduction is in the evaluation protocol: Section V-B reports that partition size 500 is chosen because it gives the lowest MAE on the synthetic dataset (Fig. 3), and then 'all other evaluations use 500 as partition size' for the same synthetic dataset whose MAE is reported as a 24.9% average improvement (Fig. 2, abstract). With no train/validation/test split stated in Section V, the headline accuracy is the optimized selection criterion rather than an independent prediction. This inflates the headline number but does not make the architecture itself tautological. The CircuitNet clustering claim rests on visual UMAP inspection rather than a quantified metric, and the KL JumpStart case study is an external application result; both are weaknesses in evidence strength, not circular definitions.

Assumptions & free parameters 1 free parameters · 4 assumptions · 0 invented entities

The model relies on a few standard results and on the modularity of VLSI graphs. The synthetic evaluation uses a self-defined ground truth, and the partition size is tuned on the same synthetic data, which ties some performance to the evaluation setup.

free parameters (1)
  • Partition size (subgraph node count) = 500
    Chosen from Fig. 3 based on lowest MAE on the synthetic dataset; this is a hand-tuned hyperparameter that affects reported performance.
assumptions (4)
  • standard math WL graph isomorphism test bounds the expressiveness of MP-GNNs
    Invoked in Section III-A to motivate transformers; cited from [27].
  • domain assumption VLSI graphs are modular, so partitioning into subgraphs preserves similarity-relevant structure
    Stated in Section IV-C and used to justify the partitioned training pipeline.
  • domain assumption The number of random edit operations is a valid ground-truth measure of graph similarity for design reuse
    Used in Section V-A1 to construct the synthetic dataset and evaluate ranking MAE.
  • domain assumption InfoGraph mutual information maximization produces useful graph embeddings
    Adopted from [24] without re-derivation; the central learning objective.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Pieceformer: Similarity-Driven Knowledge Transfer via Scalable Graph Transformer in VLSI." pith.science (2026). https://pith.science/paper/XXXPGJTU

@misc{pith2026250615907,
  author       = {Pith},
  title        = {Pith review of: Pieceformer: Similarity-Driven Knowledge Transfer via Scalable Graph Transformer in VLSI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XXXPGJTU}},
  note         = {Machine review of arXiv:2506.15907}
}
read the original abstract

Accurate graph similarity is critical for knowledge transfer in VLSI design, enabling the reuse of prior solutions to reduce engineering effort and turnaround time. We propose Pieceformer, a scalable, self-supervised similarity assessment framework, equipped with a hybrid message-passing and graph transformer encoder. To address transformer scalability, we incorporate a linear transformer backbone and introduce a partitioned training pipeline for efficient memory and parallelism management. Evaluations on synthetic and real-world CircuitNet datasets show that Pieceformer reduces mean absolute error (MAE) by 24.9% over the baseline and is the only method to correctly cluster all real-world design groups. We further demonstrate the practical usage of our model through a case study on a partitioning task, achieving up to 89% runtime reduction. These results validate the framework's effectiveness for scalable, unbiased design reuse in modern VLSI systems.

Figures

Figures reproduced from arXiv: 2506.15907 by the authors.

Figure 1
Figure 1. Overview of the proposed framework. Input graphs are partitioned into subgraphs and processed using a hybrid message-passing and partitioned [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Similarity ranking MAE across different graph scales. Five encoders are compared. Our model achieves the lowest MAE across nearly all scales, with an average improvement of 24.9% over the baseline. Transformer￾based models (GT, GIN+GT) demonstrate modest gains, while the partitioned pipeline (PGT, GIN+PGT) ensures both scalability and best performance. the number of subgraphs processed in parallel on the GPU. These … view at source ↗
Figure 3
Figure 3. Partition size impact on similarity ranking MAE. When partition size [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: 2D projection of graph embeddings on the CircuitNet dataset and Runtime analysis—illustrating cluster separability of design groups. GIN, GT, and [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 20 canonical work pages

  1. [1]

    S. G. Aksoy, K. E. Nowak, E. Purvine, and S. J. Young. Relative hausdorff distance for network analysis.Applied Network Science, 4:1– 25, 2019

    work page 2019

  2. [2]

    E. C. Barboza, N. Shukla, Y . Chen, and J. Hu. Machine learning-based pre-routing timing prediction with reduced pessimism. InProceedings of the 56th Annual Design Automation Conference 2019, pages 1–6, 2019

    work page 2019

  3. [3]

    H. Bunke. On a relation between graph edit distance and maximum common subgraph.Pattern recognition letters, 18(8):689–694, 1997

    work page 1997

  4. [4]

    Z. Chai, Y . Zhao, W. Liu, Y . Lin, R. Wang, and R. Huang. Circuitnet: An open-source dataset for machine learning in vlsi cad applications with improved domain-specific evaluation metric and learning strategies. IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2023

    work page 2023

  5. [5]

    Choromanski, V

    K. Choromanski, V . Likhosherstov, D. Dohan, X. Song, A. Gane, T. Sarlos, P. Hawkins, J. Davis, D. Belanger, L. Colwell, et al. Masked language modeling for proteins via linearly scalable long-context trans- formers.arXiv preprint arXiv:2006.03555, 2020

    arXiv 2006

  6. [6]

    W. Fang, S. Liu, H. Zhang, and Z. Xie. A self-supervised, pre-trained, and cross-stage-aligned circuit encoder provides a foundation for various design tasks. InProceedings of the 30th Asia and South Pacific Design Automation Conference, pages 505–512, 2025

    work page 2025

  7. [7]

    W. Fang, Y . Lu, S. Liu, Q. Zhang, C. Xu, L. W. Wills, H. Zhang, and Z. Xie. Transferable pre-synthesis ppa estimation for rtl designs with data augmentation techniques.IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, 2024

    work page 2024

  8. [8]

    Felippe, F

    H. Felippe, F. Battiston, and A. Kirkley. Network mutual information measures for graph similarity.Communications Physics, 7(1), Oct. 2024

    work page 2024

Show all 31 references
  1. [9]

    Z. Guo, M. Liu, J. Gu, S. Zhang, D. Z. Pan, and Y . Lin. A timing engine inspired graph neural network model for pre-routing slack prediction. InProceedings of the 59th ACM/IEEE Design Automation Conference, pages 1207–1212, 2022

    2022

  2. [10]

    Hamilton, Z

    W. Hamilton, Z. Ying, and J. Leskovec. Inductive representation learning on large graphs.Advances in neural information processing systems, 30, 2017

    2017

  3. [11]

    W. Hu, M. Fey, H. Ren, M. Nakata, Y . Dong, and J. Leskovec. Ogb-lsc: A large-scale challenge for machine learning on graphs.arXiv preprint arXiv:2103.09430, 2021

    2021 arXiv

  4. [12]

    W. Hu, M. Fey, M. Zitnik, Y . Dong, H. Ren, B. Liu, M. Catasta, and J. Leskovec. Open graph benchmark: Datasets for machine learning on graphs.Advances in neural information processing systems, 33:22118– 22133, 2020

    2020

  5. [13]

    Karypis and V

    G. Karypis and V . Kumar. A fast and high quality multilevel scheme for partitioning irregular graphs.SIAM Journal on scientific Computing, 20(1):359–392, 1998

    1998

  6. [14]

    B. W. Kernighan and S. Lin. An efficient heuristic procedure for partitioning graphs.The Bell system technical journal, 49(2):291–307, 1970

    1970

  7. [15]

    T. N. Kipf and M. Welling. Semi-supervised classification with graph convolutional networks.arXiv preprint arXiv:1609.02907, 2016

    2016 arXiv

  8. [16]

    Liang, J

    R. Liang, J. Jung, H. Xiang, L. Reddy, A. Lvov, J. Hu, and G.-J. Nam. Flowtuner: A multi-stage eda flow tuner exploiting parameter knowledge transfer. In2021 IEEE/ACM International Conference On Computer Aided Design (ICCAD), pages 1–9. IEEE, 2021

    2021

  9. [17]

    Y .-C. Lu, S. Nath, S. S. K. Pentapati, and S. K. Lim. A fast learning- driven signoff power optimization framework. InProceedings of the 39th International Conference on Computer-Aided Design, pages 1–9, 2020

    2020

  10. [18]

    Y .-C. Lu, S. S. K. Pentapati, L. Zhu, K. Samadi, and S. K. Lim. Tp-gnn: A graph neural network framework for tier partitioning in monolithic 3d ics. In2020 57th ACM/IEEE Design Automation Conference (DAC), pages 1–6. IEEE, 2020

    2020

  11. [19]

    McInnes, J

    L. McInnes, J. Healy, and J. Melville. Umap: Uniform manifold approximation and projection for dimension reduction.arXiv preprint arXiv:1802.03426, 2018

    2018 arXiv

  12. [20]

    Mirhoseini, A

    A. Mirhoseini, A. Goldie, M. Yazgan, J. Jiang, E. Songhori, S. Wang, Y .-J. Lee, E. Johnson, O. Pathak, S. Bae, et al. Chip placement with deep reinforcement learning.arXiv preprint arXiv:2004.10746, 2020

    2004 arXiv

  13. [21]

    Pearce, A

    T. Pearce, A. Brintrup, and J. Zhu. Understanding softmax confidence and uncertainty.arXiv preprint arXiv:2106.04972, 2021

    2021 arXiv

  14. [22]

    Ren and M

    H. Ren and M. Fojtik. Nvcell: Standard cell layout in advanced technology nodes with reinforcement learning. In2021 58th ACM/IEEE Design Automation Conference (DAC), pages 1291–1294. IEEE, 2021

    2021

  15. [23]

    H. Ren, G. F. Kokai, W. J. Turner, and T.-S. Ku. Paragraph: Layout parasitics and device parameter prediction using graph neural networks. In2020 57th ACM/IEEE Design Automation Conference (DAC), pages 1–6. IEEE, 2020

    2020

  16. [24]

    F.-Y . Sun, J. Hoffmann, V . Verma, and J. Tang. Infograph: Unsupervised and semi-supervised graph-level representation learning via mutual information maximization, 2020

    2020

  17. [25]

    Veli ˇckovi´c, G

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Lio, and Y . Bengio. Graph attention networks.arXiv preprint arXiv:1710.10903, 2017

    2017 arXiv

  18. [26]

    Z. Xie, R. Liang, X. Xu, J. Hu, Y . Duan, and Y . Chen. Net2: A graph attention network method customized for pre-placement net length estimation. InProceedings of the 26th Asia and South Pacific Design Automation Conference, pages 671–677, 2021

    2021

  19. [27]

    K. Xu, W. Hu, J. Leskovec, and S. Jegelka. How powerful are graph neural networks?arXiv preprint arXiv:1810.00826, 2018

    2018 arXiv

  20. [28]

    J. Xun, Z. Chai, Y . Zhao, Y . Lin, R. Wang, and R. Huang. Circuitnet 2.0: An advanced dataset for promoting machine learning innovations in realistic chip design environment. InThe Twelfth International Conference on Learning Representations, 2024

    2024

  21. [29]

    C. Ying, T. Cai, S. Luo, S. Zheng, G. Ke, D. He, Y . Shen, and T.-Y . Liu. Do transformers really perform badly for graph representation? Advances in neural information processing systems, 34:28877–28888, 2021

    2021

  22. [30]

    Zaheer, G

    M. Zaheer, G. Guruganesh, K. A. Dubey, J. Ainslie, C. Alberti, S. Ontanon, P. Pham, A. Ravula, Q. Wang, L. Yang, et al. Big bird: Transformers for longer sequences.Advances in neural information processing systems, 33:17283–17297, 2020

    2020

  23. [31]

    Zhang, H

    Y . Zhang, H. Ren, and B. Khailany. Grannite: Graph neural network inference for transferable power estimation. In2020 57th ACM/IEEE Design Automation Conference (DAC), pages 1–6. IEEE, 2020

    2020

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.