pith. sign in

arxiv: 2605.22876 · v1 · pith:IVCO452Inew · submitted 2026-05-20 · 💻 cs.LG

WeCon: An Efficient Weight-Conditioned Neural Solver for Multi-Objective Combinatorial Optimization Problems

Xuan Wu , Jinbiao Chen , Yang Li , Lijie Wen , Chunguo Wu , Yuanshu Li , Yubin Xiao , Chunyan Miao
show 2 more authors
You Zhou Di Wang
This is my paper

Pith reviewed 2026-05-25 05:59 UTC · model grok-4.3

classification 💻 cs.LG
keywords WeConneural solvermulti-objective combinatorial optimizationweight-conditionedGated Residual FusionEfficient Preference Optimizationhypervolumeinference time
0
0 comments X

The pith

WeCon solves multi-objective combinatorial optimization problems with weight-conditioned neural networks that match state-of-the-art performance at reduced inference time.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces WeCon, a neural solver for multi-objective combinatorial optimization problems that conditions on weight vectors throughout both encoding and decoding stages. It uses an encoder with attention blocks and a Gated Residual Fusion block to integrate instance features and weights, plus a Residual Fusion block in the decoder to prevent signal dilution. It also proposes Efficient Preference Optimization to generate more informative training pairs instead of random sampling. This design allows WeCon to achieve hypervolume values comparable to the current best solver while cutting inference time by 40 percent across tested problem variants.

Core claim

WeCon is an efficient weight-conditioned neural solver for MOCOPs. It features an encoder layer with three attention blocks and a Gated Residual Fusion block for harmonious interaction between instance features and weights, a plug-and-play Residual Fusion block in the decoder to alleviate weight-signal dilution, and Efficient Preference Optimization that constructs high-quality solutions for more informative training pairs. On four MOCOP variants, it matches the hypervolume of POCCO-W while reducing inference time by 40%.

What carries the argument

The Gated Residual Fusion (GRF) block in the encoder and Residual Fusion (RF) block in the decoder, which enable weight-conditioned context modeling without dilution, along with Efficient Preference Optimization (EPO) for training.

If this is right

  • WeCon can be applied to various MOCOPs like different variants across scales and distributions.
  • Neural solvers for MOCOPs can achieve SOTA performance with significantly lower inference time.
  • The EPO method improves training effectiveness by using high-quality solution pairs.
  • The fusion blocks facilitate better weight-signal integration in attention-based models.

Where Pith is reading between the lines

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

  • Similar fusion techniques could apply to other conditional optimization tasks beyond the tested MOCOPs.
  • The 40% time reduction might support real-time use in changing multi-objective settings.
  • Further tests on additional distributions would strengthen claims about broad applicability.
  • The approach may extend to different neural solver architectures.

Load-bearing premise

The four MOCOP variants and distribution patterns tested are representative of real-world use cases and the ablation studies isolate the contribution of each component without post-hoc selection bias.

What would settle it

An experiment on a new MOCOP variant or distribution pattern where WeCon no longer matches POCCO-W hypervolume or fails to deliver the 40% inference time reduction.

read the original abstract

Existing neural solvers for Multi-Objective Combinatorial Optimization Problems (MOCOPs) commonly adopt decomposition-based strategies that scalarize an MOCOP into multiple subproblems associated with distinct weight vectors. However, they either inject weights only once during decoding, limiting weight-conditioned context modeling, or primarily during encoding, causing weight-signal dilution during decoding. Moreover, preference optimization methods rely on purely random sampling to construct solution pairs for training solvers, which often produces less informative pairs and thus leads to low training effectiveness. To better address these limitations, we propose an efficient Weight-Conditioned neural solver (WeCon). Specifically, we design an encoder layer with three attention blocks and our proposed Gated Residual Fusion (GRF) block to facilitate harmonious interaction between instance features and weights, thereby generating informative weight-conditioned context. We further introduce a plug-and-play Residual Fusion (RF) block in the decoder to alleviate weight-signal dilution. Finally, we propose Efficient Preference Optimization (EPO), which constructs high-quality solutions, thereby generating more informative pairs to improve training effectiveness. Experiments on four MOCOP variants across different problem scales and distribution patterns demonstrate that WeCon achieves HyperVolume (HV) values comparable to SOTA solver POCCO-W, while reducing inference time by 40%. Ablation studies validate the contributions of all designs.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript introduces WeCon, a weight-conditioned neural solver for Multi-Objective Combinatorial Optimization Problems (MOCOPs). It identifies limitations in prior decomposition-based approaches regarding the timing of weight injection and the quality of preference optimization pairs. The method adds a Gated Residual Fusion (GRF) block to the encoder (with three attention blocks) for improved instance-weight interaction, a plug-and-play Residual Fusion (RF) block in the decoder to mitigate weight-signal dilution, and an Efficient Preference Optimization (EPO) procedure that constructs higher-quality solution pairs. Experiments across four MOCOP variants at varying scales and distributions report HyperVolume values comparable to the SOTA solver POCCO-W together with a 40% reduction in inference time; ablation studies are presented to attribute gains to each proposed component.

Significance. If the reported HV parity and inference-time gains hold under controlled conditions, the work would be a useful incremental advance in neural combinatorial optimization. The explicit GRF and RF mechanisms supply concrete, reusable patterns for maintaining weight signals across encoder-decoder stages, while EPO offers a practical improvement over random pair sampling. These elements could support faster Pareto-front approximation in time-sensitive multi-objective settings such as routing or scheduling.

major comments (1)
  1. [Abstract] Abstract and experimental results: The central claim of a 40% inference-time reduction relative to POCCO-W is load-bearing for the paper's efficiency contribution, yet no timing protocol is supplied (wall-clock per instance versus amortized batch time, hardware platform, inclusion/exclusion of data loading, or confirmation that POCCO-W was re-implemented and measured under identical conditions). Without these details the 40% figure cannot be isolated from implementation artifacts, undermining verification that GRF, RF, and EPO are responsible for the reported speedup.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for highlighting the need for a precise timing protocol to support the efficiency claims. We will revise the manuscript accordingly to enhance reproducibility and isolate the contributions of GRF, RF, and EPO.

read point-by-point responses

  1. Referee: [Abstract] Abstract and experimental results: The central claim of a 40% inference-time reduction relative to POCCO-W is load-bearing for the paper's efficiency contribution, yet no timing protocol is supplied (wall-clock per instance versus amortized batch time, hardware platform, inclusion/exclusion of data loading, or confirmation that POCCO-W was re-implemented and measured under identical conditions). Without these details the 40% figure cannot be isolated from implementation artifacts, undermining verification that GRF, RF, and EPO are responsible for the reported speedup.

    Authors: We agree that the absence of a detailed timing protocol limits the verifiability of the 40% inference-time reduction. In the revised version, we will add a new subsection under 'Experimental Setup' that specifies: (1) hardware platform (NVIDIA A100 GPU with PyTorch 2.0), (2) measurement as average wall-clock time per instance over 1000 test instances (excluding data loading and preprocessing), (3) batch size of 1 for fair single-instance comparison, and (4) explicit confirmation that POCCO-W was re-implemented from the original source code and evaluated under identical conditions on the same machine. These additions will allow readers to confirm that the speedup stems from the proposed architectural changes rather than implementation variances. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical architecture validated on external benchmarks

full rationale

The paper proposes neural architecture components (GRF block, RF block, EPO) and reports empirical results on four MOCOP variants against external SOTA solver POCCO-W. No derivation chain, equations, or fitted parameters are presented that reduce any claimed prediction or result to the inputs by construction. Performance claims rest on benchmark comparisons and ablations rather than self-referential definitions or self-citation load-bearing premises. The work is self-contained against external benchmarks with no evidence of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Based solely on the abstract, no explicit free parameters, axioms, or invented entities are enumerated; the central claim rests on empirical performance of the described neural architecture and training procedure.

pith-pipeline@v0.9.0 · 5799 in / 1064 out tokens · 24100 ms · 2026-05-25T05:59:56.022633+00:00 · methodology

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

71 extracted references · 71 canonical work pages

  1. [1]

    Reference title

    Author A, Author B, Author C. Reference title. Journal, 2024, 38: 13--28

    work page 2024

  2. [2]

    Reference title

    Author A, Author B, Author C, et al. Reference title. In: Proceedings of Conference, Place, 2024. 6--12

    work page 2024

  3. [3]

    The International Conference on Learning Representations , year=

    Neural Multi-Objective Combinatorial Optimization via Graph-Image Multimodal Fusion , author=. The International Conference on Learning Representations , year=

    work page

  4. [4]

    2026 , note=

    Efficient Neural Combinatorial Optimization Solver for the Min-max Heterogeneous Capacitated Vehicle Routing Problem , author=. 2026 , note=

    work page 2026

  5. [5]

    Reinforcement Learning-Based Nonautoregressive Solver for Traveling Salesman Problems , year=

    Xiao, Yubin and Wang, Di and Li, Boyang and Chen, Huanhuan and Pang, Wei and Wu, Xuan and Li, Hao and Xu, Dong and Liang, Yanchun and Zhou, You , journal=. Reinforcement Learning-Based Nonautoregressive Solver for Traveling Salesman Problems , year=

    work page

  6. [6]

    Science China Information Sciences , year = 2024, volume=

    Zhong, Jincheng and Ma, Haoyu and Long, Mingsheng and Wang, Jiamin , title =. Science China Information Sciences , year = 2024, volume=

    work page 2024

  7. [7]

    Science China Information Sciences , year = 2024, volume=

    Wang, Runzhong and Li, Yang and Yan, Junchi and Yang, Xiaokang , title =. Science China Information Sciences , year = 2024, volume=

    work page 2024

  8. [8]

    Science China Information Sciences , year = 2024, pages =

    Wu, Jiajun and Sun, Chengjian and Yang, Chenyang , title =. Science China Information Sciences , year = 2024, pages =

    work page 2024

  9. [9]

    The International Conference on Learning Representations , year=

    Rethinking Neural Multi-Objective Combinatorial Optimization via Neat Weight Embedding , author=. The International Conference on Learning Representations , year=

    work page

  10. [10]

    Towards Generalizable Meta-Deep Reinforcement Learning Algorithm for Multi-Objective Traveling Salesman Problems , year=

    Fu, Xiaoyu and Gu, Shenshen and Chew, Chee-Meng and Li, Tengfei , journal=. Towards Generalizable Meta-Deep Reinforcement Learning Algorithm for Multi-Objective Traveling Salesman Problems , year=

    work page

  11. [11]

    2025 , note=

    VAGPO: Vision-augmented Asymmetric Group Preference Optimization for Graph Routing Problems , author=. 2025 , note=

    work page 2025

  12. [12]

    Yubin Xiao and Di Wang and Zhiguang Cao and Rui Cao and Xuan Wu and Boyang Li and You Zhou , journal =

    work page

  13. [13]

    Liao, Zijun and Chen, Jinbiao and Wang, Debing and Zhang, Zizhen and Wang, Jiahai , booktitle =

    work page

  14. [14]

    Proceedings of the International Conference on Machine Learning , pages =

    Preference Optimization for Combinatorial Optimization Problems , author =. Proceedings of the International Conference on Machine Learning , pages =

    work page

  15. [15]

    Proceedings of the AAAI Conference on Artificial Intelligence , year =

    UCPO: A Universal Constrained Combinatorial Optimization Method via Preference Optimization , author =. Proceedings of the AAAI Conference on Artificial Intelligence , year =

    work page

  16. [16]

    Proceedings of the AAAI Conference on Artificial Intelligence , year =

    Efficient Few-Step Solution Generation via Discrete Flow Matching for Combinatorial Optimization , author =. Proceedings of the AAAI Conference on Artificial Intelligence , year =

    work page

  17. [17]

    Equity-Transformer: Solving NP-Hard Min-Max Routing Problems as Sequential Generation with Equity Context , booktitle=

    Son, Jiwoo and Kim, Minsu and Choi, Sanghyeok and Kim, Hyeonah and Park, Jinkyoo , year=. Equity-Transformer: Solving NP-Hard Min-Max Routing Problems as Sequential Generation with Equity Context , booktitle=

    work page

  18. [18]

    2025 , booktitle =

    Parallel AutoRegressive Models for Multi-Agent Combinatorial Optimization , author=. 2025 , booktitle =

    work page 2025

  19. [19]

    Proceedings of International Joint Conference on Artificial Intelligence , year = 2025, pages =

    Yubin Xiao and Yuesong Wu and Rui Cao and Di Wang and Zhiguang Cao and Xuan Wu and Peng Zhao and Yuanshu Li and You Zhou and Yuan Jiang , title =. Proceedings of International Joint Conference on Artificial Intelligence , year = 2025, pages =

    work page 2025

  20. [20]

    Proceedings of

    Kwon, Yeong-Dae and Choo, Jinho and Kim, Byoungjip and Yoon, Iljoo and Gwon, Youngjune and Min, Seungjai , year =. Proceedings of

    work page

  21. [21]

    Attention, Learn to Solve Routing Problems! , booktitle =

    Kool, Wouter and Hoof, Herke van and Welling, Max , year =. Attention, Learn to Solve Routing Problems! , booktitle =

    work page

  22. [22]

    Neural Combinatorial Optimization with Heavy Decoder:

    Luo, Fu and Li, Xin and Liu, Fei and Zhang, Qingfu and Wang, Zhenkun , year =. Neural Combinatorial Optimization with Heavy Decoder:. Proceedings of

    work page

  23. [23]

    Proceedings of International Joint Conference on Artificial Intelligence , year =

    Towards Generalizable Neural Solvers for Vehicle Routing Problems via Ensemble with Transferrable Local Policy , author =. Proceedings of International Joint Conference on Artificial Intelligence , year =

    work page

  24. [24]

    2024 , booktitle =

    Multi-Task Learning for Routing Problem with Cross-Problem Zero-Shot Generalization , author=. 2024 , booktitle =

    work page 2024

  25. [25]

    2025 , pages =

    An Efficient Diffusion-based Non-Autoregressive Solver for Traveling Salesman Problem , author=. 2025 , pages =

    work page 2025

  26. [26]

    ACM Computing Survey , articleno =

    Tian, Ye and Si, Langchun and Zhang, Xingyi and Cheng, Ran and He, Cheng and Tan, Kay Chen and Jin, Yaochu , title =. ACM Computing Survey , articleno =. 2021 , volume =

    work page 2021

  27. [27]

    2024 , booktitle =

    Wu, Yaoxin and Fan, Mingfeng and Cao, Zhiguang and Gao, Ruobin and Hou, Yaqing and Sartoretti, Guillaume , title =. 2024 , booktitle =

    work page 2024

  28. [28]

    Proceedings of IEEE Congress on Evolutionary Computation , pages=

    A Multiobjective Evolutionary Algorithm based on Decomposition and Probability Model , author=. Proceedings of IEEE Congress on Evolutionary Computation , pages=

    work page

  29. [29]

    Computers & Operations Research , volume=

    A Genetic Algorithm for A Bi-Objective Capacitated Arc Routing Problem , author=. Computers & Operations Research , volume=

    work page

  30. [30]

    2014 , journal=

    Generation of the Exact Pareto Set in Multi-Objective Traveling Salesman and Set Covering Problems , author=. 2014 , journal=

    work page 2014

  31. [31]

    Two-phase

    Lust, Thibaut and Teghem, Jacques , journal=. Two-phase

    work page

  32. [32]

    A Review on Learning to Solve Combinatorial Optimisation Problems in Manufacturing , volume =

    Zhang, Cong and Wu, Yaoxin and Ma, Yining and Song, Wen and Le, Zhang and Cao, Zhiguang and Zhang, Jie , pages =. A Review on Learning to Solve Combinatorial Optimisation Problems in Manufacturing , volume =. 2023 , journal =

    work page 2023

  33. [33]

    2024 , note=

    Neural Combinatorial Optimization Algorithms for Solving Vehicle Routing Problems: A Comprehensive Survey with Perspectives , author =. 2024 , note=

    work page 2024

  34. [34]

    A Survey on Mixture of Experts in Large Language Models , year=

    Cai, Weilin and Jiang, Juyong and Wang, Fan and Tang, Jing and Kim, Sunghun and Huang, Jiayi , journal=. A Survey on Mixture of Experts in Large Language Models , year=

    work page

  35. [35]

    2025 , booktitle =

    Liu, Qidong and Lian, Jiurui and Liu, Chaoyue and Cao, Zhiguang , title =. 2025 , booktitle =

    work page 2025

  36. [36]

    Multiobjective Combinatorial Optimization Using a Single Deep Reinforcement Learning Model , year=

    Wang, Zhenkun and Yao, Shunyu and Li, Genghui and Zhang, Qingfu , journal=. Multiobjective Combinatorial Optimization Using a Single Deep Reinforcement Learning Model , year=

    work page

  37. [37]

    2025 , booktitle =

    Li, Qi and Cao, Zhiguang and Ma, Yining and Wu, Yaoxin and Gong, Yue-Jiao , title =. 2025 , booktitle =

    work page 2025

  38. [38]

    2020 , pages =

    Hong Wu and Jiahai Wang and Zizhen Zhang , booktitle=. 2020 , pages =

    work page 2020

  39. [39]

    2022 , booktitle =

    Zong, Zefang and Wang, Hansen and Wang, Jingwei and Zheng, Meng and Li, Yong , title =. 2022 , booktitle =

    work page 2022

  40. [40]

    Deep Reinforcement Learning for Multiobjective Optimization , year=

    Li, Kaiwen and Zhang, Tao and Wang, Rui , journal=. Deep Reinforcement Learning for Multiobjective Optimization , year=

    work page

  41. [41]

    Goh, Yong Liang and Cao, Zhiguang and Ma, Yining and Zhou, Jianan and Dupty, Mohammed Haroon and Lee, Wee Sun , booktitle =

    work page

  42. [42]

    Meta-Learning-Based Deep Reinforcement Learning for Multiobjective Optimization Problems , year=

    Zhang, Zizhen and Wu, Zhiyuan and Zhang, Hang and Wang, Jiahai , journal=. Meta-Learning-Based Deep Reinforcement Learning for Multiobjective Optimization Problems , year=

    work page

  43. [43]

    Improving Pareto Local Search Using Cooperative Parallelism Strategies for Multiobjective Combinatorial Optimization , year=

    Shi, Jialong and Sun, Jianyong and Zhang, Qingfu and Zhang, Haotian and Fan, Ye , journal=. Improving Pareto Local Search Using Cooperative Parallelism Strategies for Multiobjective Combinatorial Optimization , year=

    work page

  44. [44]

    European Journal of Operational Research , volume=

    Genetic Local Search for Multi-objective Combinatorial Optimization , author=. European Journal of Operational Research , volume=

    work page

  45. [45]

    A Fast and Elitist Multiobjective Genetic Algorithm:

    Deb, Kalyanmoy and Pratap, Amrit and Agarwal, Sameer and Meyarivan, TAMT , journal=. A Fast and Elitist Multiobjective Genetic Algorithm:. 2002 , publisher=

    work page 2002

  46. [46]

    IEEE Transactions on Evolutionary Computation , volume=

    A Faster Algorithm for Calculating Hypervolume , author=. IEEE Transactions on Evolutionary Computation , volume=. 2006 , publisher=

    work page 2006

  47. [47]

    Transactions on Machine Learning Research , pages=

    RouteFinder: Towards Foundation Models for Vehicle Routing Problems , author=. Transactions on Machine Learning Research , pages=

    work page

  48. [48]

    Proceedings of International Conference on Machine Learning , pages =

    Language Modeling with Gated Convolutional Networks , author =. Proceedings of International Conference on Machine Learning , pages =

    work page

  49. [49]

    Neural Networks , volume =

    Stefan Elfwing and Eiji Uchibe and Kenji Doya , title =. Neural Networks , volume =

    work page

  50. [50]

    Proceedings of Advances in Neural Information Processing Systems , pages =

    Root Mean Square Layer Normalization , author=. Proceedings of Advances in Neural Information Processing Systems , pages =

    work page

  51. [51]

    RL4CO: An Extensive Reinforcement Learning for Combinatorial Optimization Benchmark , year =

    Berto, Federico and Hua, Chuanbo and Park, Junyoung and Luttmann, Laurin and Ma, Yining and Bu, Fanchen and Wang, Jiarui and Ye, Haoran and Kim, Minsu and Choi, Sanghyeok and Zepeda, Nayeli Gast and Hottung, Andr\'. RL4CO: An Extensive Reinforcement Learning for Combinatorial Optimization Benchmark , year =. Proceedings of ACM SIGKDD Conference on Knowled...

    work page

  52. [52]

    Zhou, Jianan and Cao, Zhiguang and Wu, Yaoxin and Song, Wen and Ma, Yining and Zhang, Jie and Chi, Xu , booktitle =

    work page

  53. [53]

    2024 , booktitle =

    Goh, Yong Liang and Cao, Zhiguang and Ma, Yining and Dong, Yanfei and Dupty, Mohammed Haroon and Lee, Wee Sun , title =. 2024 , booktitle =

    work page 2024

  54. [54]

    Machine Learning for Combinatorial Optimization:

    Bengio, Yoshua and Lodi, Andrea and Prouvost, Antoine , year =. Machine Learning for Combinatorial Optimization:. European Journal of Operational Research , volume =

    work page

  55. [55]

    Handbooks in Operations Research and Management Science , author =

    On the history of combinatorial optimization (till 1960) , volume =. Handbooks in Operations Research and Management Science , author =. 2005 , pages =

    work page 1960

  56. [56]

    2007 , volume=

    Zhang, Qingfu and Li, Hui , journal=. 2007 , volume=

    work page 2007

  57. [57]

    Proceedings of

    Neural Multi-Objective Combinatorial Optimization with Diversity Enhancement , author=. Proceedings of. 2023 , pages =

    work page 2023

  58. [58]

    International Conference on Learning Representations , year=

    Adam: A Method for Stochastic Optimization , author=. International Conference on Learning Representations , year=

    work page

  59. [59]

    SIAM Journal on Optimization , volume=

    Normal-boundary Intersection: A New Method for Generating the Pareto Surface in Nonlinear Multicriteria Optimization Problems , author=. SIAM Journal on Optimization , volume=

    work page

  60. [60]

    Conditional Neural Heuristic for Multiobjective Vehicle Routing Problems , year=

    Fan, Mingfeng and Wu, Yaoxin and Cao, Zhiguang and Song, Wen and Sartoretti, Guillaume and Liu, Huan and Wu, Guohua , journal=. Conditional Neural Heuristic for Multiobjective Vehicle Routing Problems , year=

    work page

  61. [61]

    2025 , author =

    An Optimization Framework for Multi-Objective Traffic Signal Control using Intersection Similarity and Reinforcement Learning , journal =. 2025 , author =

    work page 2025

  62. [62]

    Proceedings of

    Efficient Meta Neural Heuristic for Multi-Objective Combinatorial Optimization , author=. Proceedings of. 2023 , pages =

    work page 2023

  63. [63]

    Machine learning , volume=

    Simple statistical gradient-following algorithms for connectionist reinforcement learning , author=. Machine learning , volume=

    work page

  64. [64]

    Nie, Lei and Qi, Dandan and Liu, Bingyi and Li, Peng and Bao, Haizhou and He, Heng , journal=. CMRM:. 2025 , volume=

    work page 2025

  65. [65]

    2025 , howpublished =

    Preference-Aware Mixture-of-Experts for Multi-Objective Combinatorial Optimization , author=. 2025 , howpublished =

    work page 2025

  66. [66]

    Mathematics of Operations Research , volume=

    Proper Efficiency in Nonconvex Multicriteria Programming , author=. Mathematics of Operations Research , volume=

    work page

  67. [67]

    Pareto Set Learning for Neural Multi-Objective Combinatorial Optimization , booktitle =

    Lin, Xi and Yang, Zhiyuan and Zhang, Qingfu , year =. Pareto Set Learning for Neural Multi-Objective Combinatorial Optimization , booktitle =

    work page

  68. [68]

    Reinelt, Gerhard , journal=

    work page

  69. [69]

    Proceedings of International Conference on Machine Learning , pages =

    Han Li and Fei Liu and Zhi Zheng and Yu Zhang and Zhenkun Wang , year=. Proceedings of International Conference on Machine Learning , pages =

    work page

  70. [70]

    Zheng, Zhi and Yao, Shunyu and Wang, Zhenkun and Xialiang, Tong and Yuan, Mingxuan and Tang, Ke , booktitle =

    work page

  71. [71]

    Proceedings of

    Preference-Driven Multi-Objective Combinatorial Optimization with Conditional Computation , author=. Proceedings of

    work page