Pith. sign in

REVIEW 2 major objections 1 minor 50 references

Bi-NAS applies bi-level neural architecture search and zero-shot LLM prompting to generate personalized explanations that improve both accuracy and transparency in recommender systems.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Bi-NAS applies bi-level NAS to search explanation architectures and LLMs for text generation, reporting gains in both recommendation accuracy and explanation effectiveness across four real-world datasets.

T0 review reviewed 2026-07-03 challenge →

load-bearing objection Bi-NAS applies bi-level NAS to tune cross-attention and feature interactions then adds zero-shot LLM prompting for recsys explanations, but the abstract gives no metrics or validation protocol for the claimed explanation gains. the 2 major comments →

arxiv 2607.01387 v1 pith:LGN6HRFE submitted 2026-07-01 cs.IR cs.LG

Bi-NAS: Towards Effective and Personalized Explanation for Recommender Systems via Bi-Level Neural Architecture Search

Longfeng Wu , Yao Zhou , Tong Zeng , Zhimin Peng , Bhanu Pratap Singh Rawat , Lecheng Zheng , Giovanni Seni , Dawei Zhou This is my paper
classification cs.IR cs.LG
keywords recommender systemsneural architecture searchexplanationslarge language modelspersonalized recommendationscross-attentionfeature interactionsbi-level optimization
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 presents a Bi-NAS framework that jointly optimizes cross-attention mechanisms and feature interaction functions by searching design spaces at both intra-layer and inter-layer levels. It pairs this search with large language models that use zero-shot prompts to produce explanations aligned to user feature preferences and item quality scores. The goal is to make recommendations more accurate while giving users clearer justifications that reflect their intent and the item's attributes. If the approach holds, it would raise user trust and engagement by turning opaque suggestions into reasoned ones across varied datasets.

Core claim

Bi-NAS simultaneously refines cross-attention mechanisms and feature interaction functions by exploring both intra-layer and inter-layer design spaces, then integrates large language models through zero-shot prompting to generate explanations that align user feature preferences with item quality scores, thereby boosting recommendation accuracy and explanation effectiveness on four real-world datasets.

What carries the argument

Bi-level Neural Architecture Search (Bi-NAS) that explores intra-layer and inter-layer spaces for cross-attention and feature interaction functions, combined with zero-shot LLM prompting aligned to user features and item attributes.

Load-bearing premise

That the effectiveness of the generated explanations can be measured reliably and that the LLM outputs genuinely reflect user intent and item attributes rather than hallucination or post-hoc fitting.

What would settle it

Human ratings or quantitative metrics on four datasets showing that Bi-NAS explanations score no higher than standard post-hoc methods on clarity, reliability, or alignment with user preferences.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Recommendation accuracy increases on the evaluated real-world datasets.
  • Explanations become more effective at conveying why an item matches a user.
  • User engagement, trust, and decision-making improve because explanations are personalized and transparent.
  • Explanations better capture both user intent and item attributes through the alignment step.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same bi-level search could be tested on non-recommendation tasks that need both model performance and human-interpretable outputs.
  • Zero-shot LLM explanations might reduce the need for separate post-hoc explanation modules in production systems.
  • If the alignment between user features and item scores proves stable, it could extend to dynamic settings where preferences change over time.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The paper proposes Bi-NAS, a bi-level neural architecture search framework that jointly optimizes cross-attention mechanisms and feature interaction functions across intra- and inter-layer design spaces for recommender systems. It integrates zero-shot LLM prompting to generate personalized explanations by aligning user feature preferences with item quality scores. The central claim is that extensive evaluations on four real-world datasets show gains in both recommendation accuracy and explanation effectiveness.

Significance. If the evaluation protocol and metrics for explanation quality are rigorously defined and the results hold, the combination of bi-level NAS with LLM-based generation could advance explainable recommendation by producing more transparent, user-aligned justifications. The approach of searching both architectural components and explanation generation is a coherent extension of NAS methods to the explanation setting.

major comments (2)
  1. [Abstract] Abstract: the claim that Bi-NAS 'significantly improves the effectiveness of explanations' supplies no definition of explanation effectiveness, no fidelity or user-study metric, no baseline comparisons for explanations, and no protocol for controlling LLM hallucination. This is load-bearing for the central claim because the abstract asserts measurable gains without any concrete evaluation design.
  2. [Abstract] Abstract: no experimental protocol, statistical tests, or dataset-specific results are described for either accuracy or explanation improvements, preventing verification that the bi-level search actually produces the claimed alignment between user preferences and item attributes.
minor comments (1)
  1. The description of zero-shot prompting and feature-preference alignment would benefit from an explicit prompt template or loss term that enforces the claimed alignment.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for highlighting the need for greater precision in the abstract. We agree that the abstract should more explicitly reference the evaluation design for explanation effectiveness and the experimental protocol. We will revise the abstract accordingly in the next version.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that Bi-NAS 'significantly improves the effectiveness of explanations' supplies no definition of explanation effectiveness, no fidelity or user-study metric, no baseline comparisons for explanations, and no protocol for controlling LLM hallucination. This is load-bearing for the central claim because the abstract asserts measurable gains without any concrete evaluation design.

    Authors: We accept this observation. The current abstract is intentionally concise, but the lack of explicit definitions and metric references weakens the claim. In revision we will add a brief definition of explanation effectiveness (grounded in fidelity to user preferences and alignment with item attributes via the bi-level search), reference the fidelity and user-study metrics used, note the baseline comparisons performed, and indicate the grounding mechanism (user-feature-to-item-score alignment) employed to reduce hallucination risk. revision: yes

  2. Referee: [Abstract] Abstract: no experimental protocol, statistical tests, or dataset-specific results are described for either accuracy or explanation improvements, preventing verification that the bi-level search actually produces the claimed alignment between user preferences and item attributes.

    Authors: We agree that the abstract omits these elements. We will revise it to include a high-level statement of the evaluation protocol (four real-world datasets, standard accuracy metrics with statistical significance testing, and explanation metrics), while preserving brevity and directing readers to the detailed results and ablation studies in the body of the paper. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The provided abstract and description contain no equations, search objectives, fitted parameters, or self-citations that reduce claimed accuracy or explanation improvements to inputs by construction. The Bi-NAS framework and LLM zero-shot prompting are presented as methods whose outputs are then evaluated on datasets, with no visible self-definitional loops, renamed fits, or load-bearing self-citations. The derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The abstract alone supplies insufficient detail to enumerate concrete free parameters or axioms; the framework implicitly assumes that the chosen search space contains architectures superior to hand-designed ones and that LLM zero-shot outputs are faithful to the underlying feature scores.

axioms (2)
  • domain assumption Bi-level search over intra- and inter-layer spaces yields explanation architectures that generalize beyond the search datasets
    Invoked by the proposal that the searched designs improve effectiveness on held-out data
  • domain assumption Zero-shot LLM prompting produces justifications that accurately reflect user feature preferences and item quality scores
    Stated in the alignment step of the abstract

reviewed 2026-07-03 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Bi-NAS: Towards Effective and Personalized Explanation for Recommender Systems via Bi-Level Neural Architecture Search." pith.science (2026). https://pith.science/paper/LGN6HRFE

@misc{pith2026260701387,
  author       = {Pith},
  title        = {Pith review of: Bi-NAS: Towards Effective and Personalized Explanation for Recommender Systems via Bi-Level Neural Architecture Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LGN6HRFE}},
  note         = {Machine review of arXiv:2607.01387}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Recommender systems are vital in helping users navigate vast amounts of information, offering personalized suggestions and effective explanations for these recommendations. While previous efforts have attempted to provide such explanations, evaluating their effectiveness across various scenarios remains a challenge. Enhancing these explanations is essential for improving user engagement, trust, and decision-making. To facilitate effective explanations within the recommender system, we propose a Bi-level Neural Architecture Search (Bi-NAS) framework to optimize explanations. This approach simultaneously refines cross-attention mechanisms and feature interaction functions by exploring both intra-layer and inter-layer design spaces. Furthermore, we integrate Large Language Models (LLMs) to enhance explanation generation, leveraging zero-shot prompting to produce more effective and personalized justifications. By aligning user feature preferences with item quality scores, our approach ensures that explanations reflect both user intent and item attributes, improving transparency and reasoning depth. Extensive evaluations on four real-world datasets demonstrate that Bi-NAS not only boosts recommendation accuracy but also significantly improves the effectiveness of explanations for recommender systems, providing users with clear and reliable insights into the suggestions they receive.

Figures

Figures reproduced from arXiv: 2607.01387 by Bhanu Pratap Singh Rawat, Dawei Zhou, Giovanni Seni, Lecheng Zheng, Longfeng Wu, Tong Zeng, Yao Zhou, Zhimin Peng.

Figure 1
Figure 1. Figure 1: Illustration of recommender system with effective explanation. The [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our proposed architecture. The left part is the architecture for recommender systems, and the right part illustrates the framework of [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An overview of the optimization strategy. Each node denotes the [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Zero-shot explanation prompt template for enhancing effective and [PITH_FULL_IMAGE:figures/full_fig_p006_5.png] view at source ↗
Figure 4
Figure 4. Figure 4: A demonstration of the generated explanations for the recommendation [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Performance and explanation evaluation w/o cross-attention. [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

50 extracted references · 50 canonical work pages · 3 internal anchors

  1. [1]

    C. C. Aggarwalet al.,Recommender systems. Springer, 2016, vol. 1

    work page 2016

  2. [2]

    Towards high-order complementary recommendation via logical reasoning network,

    L. Wu, Y . Zhou, and D. Zhou, “Towards high-order complementary recommendation via logical reasoning network,” in2022 IEEE Interna- tional Conference on Data Mining (ICDM). IEEE, 2022

    work page 2022

  3. [3]

    Measuring recommendation explanation quality: The conflicting goals of explanations,

    K. Balog and F. Radlinski, “Measuring recommendation explanation quality: The conflicting goals of explanations,” inProceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval, 2020, pp. 329–338

    work page 2020

  4. [4]

    Ex3: Explainable attribute-aware item-set recommendations,

    Y . Xian, T. Zhao, J. Li, J. Chan, A. Kan, J. Ma, X. L. Dong, C. Faloutsos, G. Karypis, S. Muthukrishnanet al., “Ex3: Explainable attribute-aware item-set recommendations,” inProceedings of the 15th ACM Conference on Recommender Systems, 2021, pp. 484–494

    work page 2021

  5. [5]

    Collaborative filtering recommender systems,

    J. B. Schafer, D. Frankowski, J. Herlocker, and S. Sen, “Collaborative filtering recommender systems,” inThe adaptive web. Springer, 2007, pp. 291–324

    work page 2007

  6. [6]

    Matrix factorization techniques for recommender systems,

    Y . Koren, R. Bell, and C. V olinsky, “Matrix factorization techniques for recommender systems,”Computer, vol. 42, no. 8, pp. 30–37, 2009

    work page 2009

  7. [7]

    Asymmetrical hierarchical networks with attentive interactions for interpretable review-based recommendation,

    X. Dong, J. Ni, W. Cheng, Z. Chen, B. Zong, D. Song, Y . Liu, H. Chen, and G. De Melo, “Asymmetrical hierarchical networks with attentive interactions for interpretable review-based recommendation,” in Proceedings of the AAAI conference on artificial intelligence, vol. 34, no. 05, 2020, pp. 7667–7674

    work page 2020

  8. [8]

    Explainable recommendations via attentive multi-persona collaborative filtering,

    O. Barkan, Y . Fuchs, A. Caciularu, and N. Koenigstein, “Explainable recommendations via attentive multi-persona collaborative filtering,” in Proceedings of the 14th ACM Conference on Recommender Systems, 2020, pp. 468–473

    work page 2020

  9. [9]

    Attention-guide walk model in heterogeneous information network for multi-style recommendation explanation,

    X. Wang, Y . Wang, and Y . Ling, “Attention-guide walk model in heterogeneous information network for multi-style recommendation explanation,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 04, 2020, pp. 6275–6282

    work page 2020

  10. [10]

    Temporal meta-path guided explainable recommendation,

    H. Chen, Y . Li, X. Sun, G. Xu, and H. Yin, “Temporal meta-path guided explainable recommendation,” inProceedings of the 14th ACM international conference on web search and data mining, 2021, pp. 1056–1064

    work page 2021

  11. [11]

    Path language modeling over knowledge graphsfor explainable recommendation,

    S. Geng, Z. Fu, J. Tan, Y . Ge, G. De Melo, and Y . Zhang, “Path language modeling over knowledge graphsfor explainable recommendation,” in Proceedings of the ACM Web Conference 2022, 2022, pp. 946–955

    work page 2022

  12. [12]

    Explainable session-based recommen- dation with meta-path guided instances and self-attention mechanism,

    J. Zheng, J. Mai, and Y . Wen, “Explainable session-based recommen- dation with meta-path guided instances and self-attention mechanism,” inProceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2022

    work page 2022

  13. [13]

    A comprehensive survey of neural architecture search: Challenges and solutions,

    P. Ren, Y . Xiao, X. Chang, P.-Y . Huang, Z. Li, X. Chen, and X. Wang, “A comprehensive survey of neural architecture search: Challenges and solutions,”ACM Computing Surveys (CSUR), vol. 54, no. 4, 2021

    work page 2021

  14. [14]

    Do users rate or review? boost phrase-level sentiment labeling with review-level sentiment classification,

    Y . Zhang, H. Zhang, M. Zhang, Y . Liu, and S. Ma, “Do users rate or review? boost phrase-level sentiment labeling with review-level sentiment classification,” inProceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval, 2014, pp. 1027–1030

    work page 2014

  15. [15]

    Explicit factor models for explainable recommendation based on phrase-level sentiment analysis,

    Y . Zhang, G. Lai, M. Zhang, Y . Zhang, Y . Liu, and S. Ma, “Explicit factor models for explainable recommendation based on phrase-level sentiment analysis,” inProceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval, 2014, pp. 83–92

    work page 2014

  16. [16]

    Explainable recommendation via multi-task learning in opinionated text data,

    N. Wang, H. Wang, Y . Jia, and Y . Yin, “Explainable recommendation via multi-task learning in opinionated text data,” inThe 41st international ACM SIGIR conference on research & development in information retrieval, 2018, pp. 165–174

    work page 2018

  17. [17]

    Try this instead: Personalized and interpretable substitute recommendation,

    T. Chen, H. Yin, G. Ye, Z. Huang, Y . Wang, and M. Wang, “Try this instead: Personalized and interpretable substitute recommendation,” inProceedings of the 43rd international ACM SIGIR conference on research and development in information retrieval, 2020, pp. 891–900

    work page 2020

  18. [18]

    Counterfactual explainable recommendation,

    J. Tan, S. Xu, Y . Ge, Y . Li, X. Chen, and Y . Zhang, “Counterfactual explainable recommendation,” inProceedings of the 30th ACM Inter- national Conference on Information & Knowledge Management, 2021, pp. 1784–1793

    work page 2021

  19. [19]

    From intrinsic to counterfactual: On the explainability of contextualized recommender systems,

    Y . Zhou, H. Wang, J. He, and H. Wang, “From intrinsic to counterfactual: On the explainability of contextualized recommender systems,”arXiv preprint arXiv:2110.14844, 2021

    work page arXiv 2021

  20. [20]

    Efficient neural interaction function search for collaborative filtering,

    Q. Yao, X. Chen, J. T. Kwok, Y . Li, and C.-J. Hsieh, “Efficient neural interaction function search for collaborative filtering,” inProceedings of The web conference 2020, 2020, pp. 1660–1670

    work page 2020

  21. [21]

    A Survey on Neural Architecture Search

    M. Wistuba, A. Rawat, and T. Pedapati, “A survey on neural architecture search,”arXiv preprint arXiv:1905.01392, 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1905

  22. [22]

    Learning transferable architectures for scalable image recognition,

    B. Zoph, V . Vasudevan, J. Shlens, and Q. V . Le, “Learning transferable architectures for scalable image recognition,” inProceedings of the IEEE conference on computer vision and pattern recognition, 2018

    work page 2018

  23. [23]

    Autoloss: Automated loss function search in recommendations,

    X. Zhao, H. Liu, W. Fan, H. Liu, J. Tang, and C. Wang, “Autoloss: Automated loss function search in recommendations,” inProceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining, 2021, pp. 3959–3967

    work page 2021

  24. [24]

    Adafs: Adaptive feature selection in deep recommender system,

    W. Lin, X. Zhao, Y . Wang, T. Xu, and X. Wu, “Adafs: Adaptive feature selection in deep recommender system,” inProceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2022, pp. 3309–3317

    work page 2022

  25. [25]

    Justifying recommendations using distantly-labeled reviews and fine-grained aspects,

    J. Ni, J. Li, and J. McAuley, “Justifying recommendations using distantly-labeled reviews and fine-grained aspects,” inProceedings of the 2019 Conference on Empirical Methods in Natural Language Pro- cessing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 2019, pp. 188–197

    work page 2019

  26. [26]

    The Llama 3 Herd of Models

    A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Yang, A. Fanet al., “The llama 3 herd of models,”arXiv preprint arXiv:2407.21783, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  27. [27]

    Neural col- laborative filtering,

    X. He, L. Liao, H. Zhang, L. Nie, X. Hu, and T.-S. Chua, “Neural col- laborative filtering,” inProceedings of the 26th international conference on world wide web, 2017, pp. 173–182

    work page 2017

  28. [28]

    Vbpr: visual bayesian personalized ranking from implicit feedback,

    R. He and J. McAuley, “Vbpr: visual bayesian personalized ranking from implicit feedback,” inProceedings of the AAAI conference on artificial intelligence, vol. 30, no. 1, 2016

    work page 2016

  29. [29]

    Based explainable recom- mendations: A transparency perspective,

    Y . Zhou, H. Wang, J. He, and H. Wang, “Based explainable recom- mendations: A transparency perspective,”ACM Transactions on Recom- mender Systems, 2024

    work page 2024

  30. [30]

    Learn basic skills and reuse: Modularized adaptive neural architecture search (manas),

    H. Chen, Y . Li, H. Zhu, and Y . Zhang, “Learn basic skills and reuse: Modularized adaptive neural architecture search (manas),” in Proceedings of the 31st ACM International Conference on Information & Knowledge Management, 2022, pp. 169–179

    work page 2022

  31. [31]

    Explainable matrix factorization for col- laborative filtering,

    B. Abdollahi and O. Nasraoui, “Explainable matrix factorization for col- laborative filtering,” inProceedings of the 25th International Conference Companion on World Wide Web, 2016, pp. 5–6

    work page 2016

  32. [32]

    Co- attentive multi-task learning for explainable recommendation

    Z. Chen, X. Wang, X. Xie, T. Wu, G. Bu, Y . Wang, and E. Chen, “Co- attentive multi-task learning for explainable recommendation.” inIJCAI, 2019, pp. 2137–2143

    work page 2019

  33. [33]

    Towards reliable rare category analysis on graphs via individual calibration,

    L. Wu, B. Lei, D. Xu, and D. Zhou, “Towards reliable rare category analysis on graphs via individual calibration,” inProceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 2023, pp. 2629–2638

    work page 2023

  34. [34]

    Towards trustworthy graph neural networks and their applica- tions in recommender systems,

    L. Wu, “Towards trustworthy graph neural networks and their applica- tions in recommender systems,” in2024 IEEE International Conference on Big Data (BigData). IEEE, 2024, pp. 8250–8252

    work page 2024

  35. [35]

    Trirank: Review-aware explainable recommendation by modeling aspects,

    X. He, T. Chen, M.-Y . Kan, and X. Chen, “Trirank: Review-aware explainable recommendation by modeling aspects,” inProceedings of the 24th ACM international on conference on information and knowledge management, 2015, pp. 1661–1670

    work page 2015

  36. [36]

    Joint deep modeling of users and items using reviews for recommendation,

    L. Zheng, V . Noroozi, and P. S. Yu, “Joint deep modeling of users and items using reviews for recommendation,” inProceedings of the tenth ACM international conference on web search and data mining, 2017, pp. 425–434

    work page 2017

  37. [37]

    Transnets: Learning to transform for recommendation,

    R. Catherine and W. Cohen, “Transnets: Learning to transform for recommendation,” inProceedings of the eleventh ACM conference on recommender systems, 2017, pp. 288–296

    work page 2017

  38. [38]

    Are vision llms road- ready? a comprehensive benchmark for safety-critical driving video understanding,

    T. Zeng, L. Wu, L. Shi, D. Zhou, and F. Guo, “Are vision llms road- ready? a comprehensive benchmark for safety-critical driving video understanding,” inProceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining V . 2, 2025, pp. 5972–5983

    work page 2025

  39. [39]

    Genuine: Graph enhanced multi-level uncertainty estimation for large language models,

    T. Wang, A. Kulkarni, T. Cody, P. A. Beling, Y . Yan, and D. Zhou, “Genuine: Graph enhanced multi-level uncertainty estimation for large language models,”arXiv preprint arXiv:2509.07925, 2025

    work page arXiv 2025

  40. [40]

    Llm-generated explanations for recommender systems,

    S. Lubos, T. N. T. Tran, A. Felfernig, S. Polat Erdeniz, and V .-M. Le, “Llm-generated explanations for recommender systems,” inAdjunct Proceedings of the 32nd ACM Conference on User Modeling, Adaptation and Personalization, 2024, pp. 276–285

    work page 2024

  41. [41]

    Neural architecture search with reinforcement learning,

    B. Zoph and Q. V . Le, “Neural architecture search with reinforcement learning,” in5th International Conference on Learning Representations, ICLR 2017, Toulon, France, April 24-26, 2017, Conference Track Proceedings, 2017

    work page 2017

  42. [42]

    Efficient neural architecture search via parameters sharing,

    H. Pham, M. Guan, B. Zoph, Q. Le, and J. Dean, “Efficient neural architecture search via parameters sharing,” inInternational conference on machine learning. PMLR, 2018, pp. 4095–4104

    work page 2018

  43. [43]

    Block-wisely supervised neural architecture search with knowledge distillation,

    C. Li, J. Peng, L. Yuan, G. Wang, X. Liang, L. Lin, and X. Chang, “Block-wisely supervised neural architecture search with knowledge distillation,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020, pp. 1989–1998

    work page 2020

  44. [44]

    DARTS: Differentiable Architecture Search

    H. Liu, K. Simonyan, and Y . Yang, “Darts: Differentiable architecture search,”arXiv preprint arXiv:1806.09055, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  45. [45]

    Autofield: Automating feature selection in deep recommender systems,

    Y . Wang, X. Zhao, T. Xu, and X. Wu, “Autofield: Automating feature selection in deep recommender systems,” inProceedings of the ACM Web Conference 2022, 2022, pp. 1977–1986

    work page 2022

  46. [46]

    Autocross: Automatic feature crossing for tabular data in real-world applications,

    Y . Luo, M. Wang, H. Zhou, Q. Yao, W.-W. Tu, Y . Chen, W. Dai, and Q. Yang, “Autocross: Automatic feature crossing for tabular data in real-world applications,” inProceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2019

    work page 2019

  47. [47]

    Autofis: Automatic feature interaction selection in factorization models for click-through rate prediction,

    B. Liu, C. Zhu, G. Li, W. Zhang, J. Lai, R. Tang, X. He, Z. Li, and Y . Yu, “Autofis: Automatic feature interaction selection in factorization models for click-through rate prediction,” inproceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining, 2020, pp. 2636–2645

    work page 2020

  48. [48]

    Nasrec: weight sharing neural architecture search for recommender systems,

    T. Zhang, D. Cheng, Y . He, Z. Chen, X. Dai, L. Xiong, F. Yan, H. Li, Y . Chen, and W. Wen, “Nasrec: weight sharing neural architecture search for recommender systems,” inProceedings of the ACM Web Conference 2023, 2023, pp. 1199–1207

    work page 2023

  49. [49]

    Rankitect: Ranking architecture search battling world-class engineers at meta scale,

    W. Wen, K.-H. Liu, I. Fedorov, X. Zhang, H. Yin, W. Chu, K. Hassani, M. Sun, J. Liu, X. Wanget al., “Rankitect: Ranking architecture search battling world-class engineers at meta scale,”arXiv preprint arXiv:2311.08430, 2023

    work page arXiv 2023

  50. [50]

    Automl for large capacity modeling of meta ranking systems,

    H. Yin, K.-H. Liu, M. Sun, Y . Chen, B. Zhang, J. Liu, V . Sehgal, R. R. Panchal, E. Hotaj, X. Liuet al., “Automl for large capacity modeling of meta ranking systems,”arXiv preprint arXiv:2311.07870, 2023

    work page arXiv 2023

This paper was first reviewed by grok-4.3 on July 3, 2026.