Pith. sign in

REVIEW 5 major objections 5 minor 41 references

Transferrable Surrogates in Expressive Neural Architecture Search Spaces

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 This paper claims that a fine-tuned language model can act as a surrogate for architecture accuracy in a flexible grammar-based search space, transferring across datasets and steering evolutionary search to stronger final models in far…

desk verdict Useful empirical study of LM surrogates in einspace, but the search claims outrun the evidence: Table 4 contradicts 'consistently' and the transfer-RF results rest on imputed runs. read the letter →

arxiv 2504.12971 v3 pith:GKII5KLL submitted 2025-04-17 cs.LG cs.AI

Shiwen Qin , Gabriela Kadlecová , Martin Pilát , Shay B. Cohen , Roman Neruda , Elliot J. Crowley , Jovita Lukasik , Linus Ericsson This is my paper · ORCID
classification cs.LGcs.AI
keywords neuralarchitecturesearchsurrogatemodelslanguagegrammar-basedspaceevolutionaryzero-costproxiesperformancepredictiontransferlearning
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 is trying to establish that cheap surrogate models can make neural architecture search (NAS) practical in large, expressive search spaces defined by context-free grammars, where previous predictors designed for small cell-based spaces do not transfer. The central result is that a fine-tuned ModernBERT-large language model, fed the string form of an architecture's derivation tree enriched with output tensor shapes, predicts validation accuracy with Kendall rank correlation around 0.61 on CIFAR-10 and transfers to the unseen NAS benchmark tasks, reaching 0.658 on Gutenberg in leave-one-out experiments. These predictions are strong enough to guide regularised evolution: surrogate-guided search reaches the baseline's best accuracy in far fewer iterations and, on most of the nine datasets, ends with higher final accuracy than the unguided baseline. The paper further claims the surrogate can be used directly as the search objective, with a large speed-up and competitive results on some tasks. If right, this would remove the need to fully train most candidate architectures during search in grammar-based spaces.

What carries the argument

The load-bearing object is the derivation-tree string encoding of an architecture: a bracket-structured text that spells out the branching, routing, aggregation and computation operations generated by the context-free grammar, optionally annotated with the output tensor shape of each operation. This string is what a language model is fine-tuned on, using a mean-squared-error loss against true validation accuracy; the shape annotations matter, since they lift Kendall correlation from 0.448 to 0.612 in the paper's ablation. For the non-LM baseline, the machinery is a descriptor concatenating zero-cost proxy scores (grad_norm, snip, grasp, fisher, jacob_cov, plain, synflow) with GRAF topological features, fed to a random forest or gradient-boosted tree regressor. The surrogate is inserted into regularised evolution at two points: selecting the top-k offspring to fully evaluate, and optionally replacing the evaluation objective entirely.

What would settle it

A decisive test is to search for string collisions: two architectures whose derivation-tree strings and shape annotations are identical but whose compiled networks differ, for example through stochastic training or hidden implementation choices in the grammar's operators. If the true validation accuracies of such pairs differ substantially, the encoding is under-informative and the surrogate cannot be a reliable predictor, since it has no way to distinguish them. A second check is leave-one-out transfer to a task whose architectures use operation statistics far outside the training distributions, such as a very different image resolution or sequence length; a collapse in rank correlation would indicate that the transfer claim is limited to aligned tasks.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that the derivation-tree string representation of architectures in einspace is a usable substrate for performance prediction. ModernBERT-large fine-tuned on these strings with output-shape metadata achieves the highest rank correlations among all tested predictors (Spearman 0.745, Kendall 0.612 on held-out CIFAR-10 architectures; 0.769/0.628 with data augmentation), clearly above a random forest trained on zero-cost proxy and GRAF graph features (0.663/0.514). The same model generalises across datasets, with leave-one-out Kendall correlations up to 0.658 on Gutenberg and 0.625 on MultNIST, and it outperforms plain few-shot prompting with open LLMs. In search, using the surrogate to select which mutated offspring to train (sampling k=5 from n=20 candidates) improves final accuracy over baseline evolution on most of the Unseen NAS datasets, and the surrogate-as-objective variant is competitive on some datasets while skipping most full evaluations.

Load-bearing premise

The paper's strongest results rest on the assumption that the derivation-tree string plus output tensor shapes is a faithful encoding of everything that determines an architecture's accuracy; if two functionally different architectures can share the same string, or if the string omits a decisive detail, the language-model surrogate cannot be right, no matter how well it is trained.

Editorial extensions

If this is right

  • Surrogate-guided search can reach the accuracy of the unguided baseline in many fewer iterations, so the wall-clock cost of exploring a grammar-defined space drops sharply.
  • A surrogate trained once on several datasets can be applied zero-shot to a new task, meaning a new NAS run may not need an initial fully-evaluated population of architectures, instead using transferred predictions to filter candidates.
  • Using the surrogate directly as the objective removes almost all full training from the search loop, trading final accuracy for an order-of-magnitude speed-up and offering a cheap way to initialise or prune search populations.
  • The gap between the best surrogate (fine-tuned ModernBERT-large) and the random forest baseline shows that learned encodings capture more of what determines accuracy than hand-engineered graph and cost features in expressive spaces.

Reading between the lines

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

  • We infer that the output-shape annotations act as a disambiguating signal for tensor flow, telling the language model how information enters and leaves each module; without it, the string under-determines the computation.
  • We infer that the same string-surrogate approach should extend to other grammar-defined search spaces, since only the grammar and the string encoder need to change, though the paper does not test this.
  • We infer that the Isabella and Chesseract results reveal a domain-alignment bottleneck: surrogates transfer well within a modality but struggle when architecture statistics differ sharply, so a practical system would likely need per-modality surrogates.
  • We infer that the augmentation scheme (reordering branches, identity insertions, shape perturbation) acts as an implicit regulariser and could be pushed further with contrastive or variational string objectives.
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. The paper studies surrogate models for neural architecture search in the grammar-based einspace search space. It proposes two families of surrogates: a random forest trained on zero-cost proxy and GRAF features, and a fine-tuned ModernBERT language model trained on string representations of architecture derivation trees augmented with output-shape metadata. The authors report rank correlations within and across datasets, evaluate surrogate-guided regularized evolution, and also test using the surrogate directly as the search objective. Their central claims are that the surrogates have high predictive power, that they transfer across datasets, that they can filter out bad architectures when searching on novel datasets, and that they can be used as a standalone search objective for large speed-ups.

Significance. If the claims hold, this would be a valuable step toward efficient NAS in large, expressive, grammar-based search spaces where fixed cell encodings do not apply. The paper is the first to evaluate a broad set of performance predictors in einspace, and the LM-based surrogate on derivation-tree strings is a novel and interesting idea. The authors are transparent about standard errors and explicitly list limitations, and the appendix discloses a bug that affects the transfer-RF experiments. However, the strength of the headline claims currently exceeds what the evidence supports, and the main transfer-related search result is explicitly labeled as not final. The central idea is plausible and worth publishing after a substantive revision that brings the claims in line with the data.

major comments (5)
  1. [Abstract; Section 4.1, Table 4] The claim that surrogate-guided evolution 'consistently achieves stronger final architecture performances' is contradicted by Table 4. Evolution(BERT) is numerically worse than baseline on MultNIST (0.765 vs 0.841) and CIFARTile (0.341 vs 0.358), and the apparent gains on GeoClassing (0.735 vs 0.725) and Chesseract (0.606 vs 0.595) are within one standard error; several other differences (e.g., AddNIST, Language) are also within one standard error. With only three seeds per condition, the evidence supports 'competitive or better on some tasks', not 'consistently stronger final architecture performances'.
  2. [Appendix B.4 footnote; Table 4 'Evolution(RF Transfer)'] The transfer-RF search results are presented as final despite the footnote in Appendix B.4 stating that an inconsistency in synflow scaling was found and fixed only shortly before the deadline, that 7 of 27 transfer experiments were not re-run to 300 iterations, and that the remaining iterations were imputed using the best value found so far. The authors write that they 'will update the results for the final version of the paper.' Since claim (ii) about filtering out bad architectures on novel datasets rests on this row, the evidence is unverified as submitted. The affected experiments should be re-run at full length, or the corresponding claims should be narrowed or marked preliminary.
  3. [Section 4.1, Tables 2 and 3] The text states that 'ModernBERT-large consistently outperforms the random forest model', but Table 3 shows the random forest with higher Spearman/Kendall correlations on Isabella (0.308/0.211 vs 0.273/0.186), GeoClassing (0.693/0.504 vs 0.661/0.475), and Chesseract (0.599/0.423 vs 0.545/0.384). The word 'consistently' should be removed, and the description should acknowledge that the LM advantage is an average trend rather than a per-dataset guarantee.
  4. [Section 3.4; Section 4.1, Table 4] The claim that surrogates 'can be used to filter out bad architectures when searching on novel datasets' is tested in search only with the random-forest transfer surrogate (Evolution(RF Transfer)). No search experiment uses a transfer-trained ModernBERT-large surrogate; the LM transfer evidence is limited to static correlation tables (Tables 2 and 3). As a result, the transfer capability of the paper's proposed best method is untested in the search setting. Either add a transfer-LM search experiment or limit the transfer-search claim to the RF variant.
  5. [Appendix B.2; Table 1] The paper states that the best LM checkpoint is saved based on Kendall Tau correlation on the evaluation set, but Table 1 reports correlations on an evaluation set for the same CIFAR10 data. Please clarify whether the reported evaluation set is disjoint from the checkpoint-selection set; if the same data are used for both checkpoint selection and final reporting, the absolute correlations are optimistically biased and model comparisons may be unfair.
minor comments (5)
  1. [Table 4] The header contains the typo 'Langauge' instead of 'Language'.
  2. [Appendix A.1, item 4] The sentence 'This augmentation changes the the architecture functionally' contains a duplicated article; it should read 'changes the architecture functionally'.
  3. [Figure 2 caption] The caption ends with a stray '>' character after 'baseline'; it should read 'outperforms the baseline.'
  4. [Table 2] The table caption does not explain the slash-separated format (e.g., '0.612/0.648'); please add a note stating which number belongs to ModernBERT-large and which to the random forest.
  5. [Section 3.2] The phrase 'we restrict our analysis to a single operation type' is vague; please specify which operation type was used and how it was selected for the node-degree features.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: surrogates are empirically fitted and evaluated on held-out data; self-citations are external benchmarks, not load-bearing.

full rationale

The paper's central claims are empirical: surrogate models are trained on architecture-accuracy pairs and evaluated on held-out architectures (Tables 1-3), then used inside an evolutionary loop whose final architectures are still fully trained and evaluated (Table 4). There is no derivation chain in which an output is identified with an input by construction. The LM surrogate's encoding is ablated in Appendix C.1, and the best checkpoint is selected on a validation set (Appendix B.2), which is standard model selection rather than fitting the reported test numbers. The paper's self-citations, einspace (Ericsson et al., 2024) and GRAF (Kadlecová et al., 2024), are externally published benchmarks and methods, and they supply the search space and feature extractor rather than justifying the surrogate's predictive validity; neither is invoked as a uniqueness theorem. The Appendix B.4 footnote discloses an inconsistency in the transfer-RF search and imputation of seven truncated runs; this weakens the strength of the transfer claims and the 'consistently' wording relative to Table 4, but it is an evidence-quality issue, not circular reasoning. No step reduces to the quantity being predicted, and no fitted parameter is renamed as a prediction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

No new physical or conceptual entities are introduced. The central claims rest on standard machine-learning assumptions about feature informativeness and the sufficiency of string encodings, plus several hand-set experimental hyperparameters that affect the magnitude but not the qualitative direction of the results.

free parameters (3)
  • k (number of selected offspring per iteration) = 5
    Hand-set compromise in the evolutionary algorithm; surrogate variants and random baseline all evaluate 5 full architectures per iteration, so search speed-up claims are relative to iterations, not wall-clock.
  • n (mutated candidates per iteration) = 20
    Hand-set to balance diversity and mutation cost; affects the pool from which the surrogate selects the top-k.
  • augmentation label noise standard deviation = 0.005
    Gaussian noise added to labels of augmented architectures; hand-set to preserve label realism while adding diversity.
assumptions (4)
  • domain assumption Zero-cost proxies and graph features computed on a network correlate with its post-training accuracy.
    Foundation of the RF/XGB surrogate; supported empirically in prior work (Abdelfattah et al. 2021, Kadlecova et al. 2024) and by Table 7 in this paper, which shows weak but nonzero correlations for most ZCPs.
  • domain assumption A pretrained language model can learn the mapping from architecture derivation strings to performance through fine-tuning.
    Central premise of the LM surrogate; supported by the paper's correlation results but not proven independently.
  • domain assumption The derivation-tree string with shape metadata is a sufficient and faithful architecture encoding.
    Ablation in Appendix C.1 shows it outperforms alternatives, but it remains an assumed sufficient representation for performance prediction.
  • domain assumption Percentile normalization across datasets makes accuracy values comparable for joint surrogate training.
    Used in leave-one-out transfer experiments; Table 6 shows it performs similarly to min-max, but comparability is assumed.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Transferrable Surrogates in Expressive Neural Architecture Search Spaces." pith.science (2026). https://pith.science/paper/GKII5KLL

@misc{pith2026250412971,
  author       = {Pith},
  title        = {Pith review of: Transferrable Surrogates in Expressive Neural Architecture Search Spaces},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GKII5KLL}},
  note         = {Machine review of arXiv:2504.12971}
}
read the original abstract

Neural architecture search (NAS) faces a challenge in balancing the exploration of expressive, broad search spaces that enable architectural innovation with the need for efficient evaluation of architectures to effectively search such spaces. We investigate surrogate model training for improving search in highly expressive NAS search spaces based on context-free grammars. We show that i) surrogate models trained either using zero-cost-proxy metrics and neural graph features (GRAF) or by fine-tuning an off-the-shelf LM have high predictive power for the performance of architectures both within and across datasets, ii) these surrogates can be used to filter out bad architectures when searching on novel datasets, thereby significantly speeding up search and achieving better final performances, and iii) the surrogates can be further used directly as the search objective for huge speed-ups.

Figures

Figures reproduced from arXiv: 2504.12971 by the authors.

Figure 1
Figure 1. The encodings used by our language models (left) and our random forest and XGBoost models (right). A derivation tree is obtained from the grammar by sampling or mutation, and encoded into a string representation. The derivation tree can be compiled into a neural network, from which we can extract ZCP scores and GRAF encodings that are ultimately concatenated to form a descriptor for the RF and XGBoost models. with a… view at source ↗
Figure 2
Figure 2. Search results using different surrogates. We plot the validation accuracies of the best models at each iteration of search, with the mean and standard error across three seeds. ModernBERT: 396M parameters). We fine-tune both size variants of these models using the string representations of architectures from their derivation trees in einspace, as described in Section 3.3. Open-source LLM: We additionally use Llama3… view at source ↗
Figure 3
Figure 3. We apply four forms of augmentation to the architecture dataset to increase the effective data size for LM training. Three of our augmentations preserve the same exact network, while the last causes minor changes: (1) We swap the nesting order of sequential models; (2) We swap the two branches in a branching(2) module; (3) We add components that reduce to the identity operation; and (4) We change the output dimensio… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Extended runs for ModernBERT-large surrogates. [PITH_FULL_IMAGE:figures/full_fig_p018_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 31 canonical work pages

  1. [1]

    S., Mehrotra, A., Dudziak, ., and Lane, N

    Abdelfattah, M. S., Mehrotra, A., Dudziak, ., and Lane, N. D. (2021). Zero-Cost Proxies for Lightweight NAS . In International Conference on Learning Representations (ICLR)

    work page 2021

  2. [2]

    and Abdelfattah, M

    Akhauri, Y. and Abdelfattah, M. S. (2024). Encodings for prediction-based neural architecture search. In Forty-first International Conference on Machine Learning, ICML

    work page 2024

  3. [3]

    Chen, A., Dohan, D., and So, D. (2023). Evoprompting: Language models for code-level neural architecture search. Advances in neural information processing systems , 36:7787--7817

    work page 2023

  4. [4]

    Chen, L., Xu, F., Li, N., Han, Z., Wang, M., Li, Y., and Hui, P. (2024). Large language model-driven meta-structure discovery in heterogeneous information network. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages 307--318

    work page 2024

  5. [5]

    Chen, M., Peng, H., Fu, J., and Ling, H. (2021a). Autoformer: Searching transformers for visual recognition. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) , pages 12270--12280

    work page 2021

  6. [6]

    Chen, M., Wu, K., Ni, B., Peng, H., Liu, B., Fu, J., Chao, H., and Ling, H. (2021b). Searching the search space of vision transformer. Advances in Neural Information Processing Systems , 34:8714--8726

    work page 2021

  7. [7]

    and Yang, Y

    Dong, X. and Yang, Y. (2020). Nas-bench-201: Extending the scope of reproducible neural architecture search. In International Conference on Learning Representations (ICLR)

    work page 2020

  8. [8]

    Dudziak, L., Chau, T., Abdelfattah, M., Lee, R., Kim, H., and Lane, N. (2020). Brp-nas: Prediction-based nas using gcns. Advances in neural information processing systems , 33:10480--10490

    work page 2020

Show all 41 references
  1. [9]

    Erickson, N., Mueller, J., Shirkov, A., Zhang, H., Larroy, P., Li, M., and Smola, A. (2020). Autogluon-tabular: Robust and accurate automl for structured data. arXiv preprint arXiv:2003.06505

    2020 arXiv

  2. [10]

    J., Cohen, S

    Ericsson, L., Espinosa, M., Yang, C., Antoniou, A., Storkey, A. J., Cohen, S. B., McDonagh, S., and Crowley, E. J. (2024). einspace: Searching for neural architectures from fundamental operations. In Advances in Neural Information Processing Systems 38: Annual Conference on Ne...

    2024

  3. [11]

    Geada, R., Towers, D., Forshaw, M., Atapour-Abarghouei, A., and McGough, A. S. (2024). Insights from the use of previously unseen neural architecture search datasets. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 22541--22550

    2024

  4. [12]

    Grattafiori, A., Dubey, A., Jauhri, A., Pandey, A., Kadian, A., Al-Dahle, A., Letman, A., Mathur, A., Schelten, A., Vaughan, A., et al. (2024). The llama 3 herd of models. arXiv preprint arXiv:2407.21783

    2024 arXiv

  5. [13]

    Jawahar, G., Abdul-Mageed, M., Lakshmanan, L. V. S., and Ding, D. (2024). Llm performance predictors are good initializers for architecture search

    2024

  6. [14]

    Kadlecov \' a , G., Lukasik, J., Pil \' a t, M., Vidnerov \' a , P., Safari, M., Neruda, R., and Hutter, F. (2024). Surprisingly strong performance prediction with neural graph features. In Forty-first International Conference on Machine Learning, ICML . OpenReview.net

    2024

  7. [15]

    Krishnakumar, A., White, C., Zela, A., Tu, R., Safari, M., and Hutter, F. (2022). NAS -bench-suite-zero: Accelerating research on zero cost proxies. In Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    2022

  8. [16]

    Krizhevsky, A. (2009). Learning multiple layers of features from tiny images. Technical report

    2009

  9. [17]

    Li, C., Tang, T., Wang, G., Peng, J., Wang, B., Liang, X., and Chang, X. (2021). BossNAS : Exploring Hybrid CNN -transformers with Block -wisely Self -supervised Neural Architecture Search . 2021 IEEE/CVF International Conference on Computer Vision (ICCV) , pages 12261--12271....

    2021

  10. [18]

    Liu, H., Simonyan, K., and Yang, Y. (2019a). DARTS : Differentiable architecture search. In International Conference on Learning Representations

    2019

  11. [19]

    Liu, Y., Ott, M., Goyal, N., Du, J., Joshi, M., Chen, D., Levy, O., Lewis, M., Zettlemoyer, L., and Stoyanov, V. (2019b). Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692

    2019 arXiv

  12. [20]

    Lukasik, J., Moeller, M., and Keuper, M. (2024). An evaluation of zero-cost proxies -- from neural architecture performance to model robustness. In International Journal of Computer Vision

    2024

  13. [21]

    Mellor, J., Turner, J., Storkey, A., and Crowley, E. J. (2021). Neural architecture search without training. In ICML

    2021

  14. [22]

    U., Earle, S., Togelius, J., James, S., and Cleghorn, C

    Nasir, M. U., Earle, S., Togelius, J., James, S., and Cleghorn, C. (2024). Llmatic: neural architecture search via large language models and quality diversity optimization. In proceedings of the Genetic and Evolutionary Computation Conference , pages 1110--1118

    2024

  15. [23]

    Ning, X., Tang, C., Li, W., Zhou, Z., Liang, S., Yang, H., and Wang, Y. (2021). Evaluating efficient performance estimators of neural architectures. In Advances in Neural Information Processing Systems

    2021

  16. [24]

    Ning, X., Zheng, Y., Zhou, Z., Zhao, T., Yang, H., and Wang, Y. (2023). A generic graph-based neural architecture encoding scheme with multifaceted information. IEEE Trans. Pattern Anal. Mach. Intell. , 45(7):7955--7969

    2023

  17. [25]

    GPT-4 Technical Report

    OpenAI (2023). GPT-4 Technical Report

    2023

  18. [26]

    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E. (2011). Scikit-learn: Machine learning in P ython. Jou...

    2011

  19. [27]

    Real, E., Aggarwal, A., Huang, Y., and Le, Q. V. (2019). Regularized evolution for image classifier architecture search. In Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence and Thirty-First Innovative Applications of Artificial Intelligence Conference...

    2019

  20. [28]

    S., Brox, T., and Hutter, F

    Schrodi, S., Stoll, D., Ru, B., Sukthanker, R. S., Brox, T., and Hutter, F. (2023). Construction of hierarchical neural architecture search spaces based on context-free grammars. In Thirty-seventh Conference on Neural Information Processing Systems

    2023

  21. [29]

    W., Parnichkun, R., Amini, A., Massaroli, S., and Poli, M

    Thomas, A. W., Parnichkun, R., Amini, A., Massaroli, S., and Poli, M. (2025). STAR : Synthesis of Tailored Architectures . In ICLR

    2025

  22. [30]

    Velickovic, P., Cucurull, G., Casanova, A., Romero, A., Li \` o , P., and Bengio, Y. (2018). Graph attention networks. In 6th International Conference on Learning Representations, ICLR

    2018

  23. [31]

    Wang, H., Gao, Y., Zheng, X., Zhang, P., Chen, H., Bu, J., and Yu, P. S. (2023). Graph neural architecture search with gpt-4. arXiv preprint arXiv:2310.01436

    2023

  24. [32]

    Warner, B., Chaffin, A., Clavi \'e , B., Weller, O., Hallstr \"o m, O., Taghadouini, S., Gallagher, A., Biswas, R., Ladhak, F., Aarsen, T., et al. (2024). Smarter, better, faster, longer: A modern bidirectional encoder for fast, memory efficient, and long context finetuning an...

    2024 arXiv

  25. [33]

    White, C., Safari, M., Sukthanker, R., Ru, B., Elsken, T., Zela, A., Dey, D., and Hutter, F. (2023). Neural architecture search: Insights from 1000 papers

    2023

  26. [34]

    White, C., Zela, A., Ru, R., Liu, Y., and Hutter, F. (2021). How powerful are performance predictors in neural architecture search? Advances in Neural Information Processing Systems , 34

    2021

  27. [35]

    Yan, S., White, C., Savani, Y., and Hutter, F. (2021). Nas-bench-x11 and the power of learning curves. In Advances in Neural Information Processing Systems

    2021

  28. [36]

    Yan, S., Zheng, Y., Ao, W., Zeng, X., and Zhang, M. (2020). Does unsupervised architecture representation learning help neural architecture search? In Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems

    2020

  29. [37]

    Yang, A., Yang, B., Zhang, B., Hui, B., Zheng, B., Yu, B., Li, C., Liu, D., Huang, F., Wei, H., et al. (2024). Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115

    2024 arXiv

  30. [38]

    Ying, C., Klein, A., Christiansen, E., Real, E., Murphy, K., and Hutter, F. (2019). NAS -bench-101: Towards reproducible neural architecture search. In Chaudhuri, K. and Salakhutdinov, R., editors, Proceedings of the 36th International Conference on Machine Learning , volume 9...

    2019

  31. [39]

    Yu, C., Liu, X., Wang, Y., Liu, Y., Feng, W., Deng, X., Tang, C., and Lv, J. (2023). Gpt-nas: Evolutionary neural architecture search with the generative pre-trained model. arXiv preprint arXiv:2305.05351

    2023 arXiv

  32. [40]

    N., Zimmer, L., Lukasik, J., Keuper, M., and Hutter, F

    Zela, A., Siems, J. N., Zimmer, L., Lukasik, J., Keuper, M., and Hutter, F. (2020). Surrogate nas benchmarks: Going beyond the limited search spaces of tabular nas benchmarks. In International Conference on Learning Representations

    2020

  33. [41]

    Zhou, X., Wu, X., Feng, L., Lu, Z., and Tan, K. C. (2024). Design principle transfer in neural architecture search via large language models

    2024

Pith tools

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