avatar

Shengyu Feng

Ph.D. Student
Carnegie Mellon University
shengyuf (at) cs.cmu.edu


About Me

I am a final-year Ph.D. student in CMU LTI, advised by Prof. Yiming Yang.

Before that, I received my M.S. in Computer Science from UIUC (2022), B.S.E in Computer Science from UMich (2020), and B.S in Electrical and Computer Engineering from Shanghai Jiao Tong University (2020).

I study learning and inference algorithms for sequential search and decision making in discrete domains, with applications to language modeling, combinatorial optimization, and scientific discovery. My research broadly spans generative models, reinforcement learning, and probabilistic inference.

Research Themes

Discrete Optimization

Learning-based methods for solving NP-hard problems, including combinatorial optimization, stochastic search, and inference-time algorithms.

Generative Models

Discrete generative modeling, probabilistic inference, and trajectory-based learning methods for reasoning and optimization.

Reinforcement Learning

Sequential decision making, search, and adaptive optimization under long-horizon environments.

AI for Mathematics & Science

AI systems for mathematical reasoning, scientific workflows, and large-scale scientific discovery.

News

Selected Publications

Unsupervised Diffusion Solver for Combinatorial Optimization via Combinatorial Adjoint Matching ICML
Unsupervised Diffusion Solver for Combinatorial Optimization via Combinatorial Adjoint Matching
Shengyu Feng, Tarun Suresh, and Yiming Yang
International Conference on Machine Learning (ICML), 2026.
First adjoint framework in discrete space for unsupervised diffusion model training.
HTML Code
FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization ICLR
FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization
Shengyu Feng*, Weiwei Sun*, Shanda Li, Ameet Talwalker, and Yiming Yang
International Conference on Learning Representations (ICLR), 2026.
Unified evaluation of both neural solvers and LLM agentic solvers on real-world & large-scale CO problems.
HTML Code
Regularized Langevin Dynamics for Combinatorial Optimization ICML
Regularized Langevin Dynamics for Combinatorial Optimization
Shengyu Feng and Yiming Yang
International Conference on Machine Learning (ICML), 2025.
Novel dynamics to address the local optima in combinatorial space, 97% faster than SOTA
HTML Code
Step-by-Step Reasoning for Math Problems via Twisted Sequential Monte Carlo ICLR
Step-by-Step Reasoning for Math Problems via Twisted Sequential Monte Carlo
Shengyu Feng, Xiang Kong, Shuang Ma, Aonan Zhang, Dong Yin, Chong Wang, Ruoming Pang, and Yiming Yang
International Conference on Learning Representations (ICLR), 2025.
LLM inference-scaling via twisted sequential Monte Carlo; unified probabilistic lens for verification methods.
HTML Code
Complete Publication List
  1. ICML
    Unsupervised Diffusion Solver for Combinatorial Optimization via Combinatorial Adjoint Matching
    Shengyu Feng, Tarun Suresh, and Yiming Yang
    International Conference on Machine Learning (ICML), 2026.
    HTML Code
  2. ICML
    Unsupervised Neural Langevin Sampler for Mixed Integer Linear Programming
    Yixin Huang, Shengyu Feng, and Yiming Yang
    International Conference on Machine Learning (ICML), 2026.
  3. ACL
    AdvancedIF: Rubric-Based Benchmarking and Reinforcement Learning for Advancing LLM Instruction Following
    Yun He, Wenzhe Li, Hejia Zhang, Songlin Li, Karishma Mandyam, Sopan Khosla, Yuanhao Xiong, Nanshu Wang, Xiaoliang Peng, Beibin Li, Shengjie Bi, Shishir G Patil, Qi Qi, Shengyu Feng, Julian Katz-Samuels, Richard Yuanzhe Pang, Sujan Kumar Gonugondla, Hunter Lang, Yue Yu, Yundi Qian, Maryam Fazel-Zarandi, Licheng Yu, Amine Benhalloum, Hany Hassan Awadalla, Manaal Faruqui
    Annual Meeting of the Association for Computational Linguistics (ACL), 2026.
    HTML
  4. arXiv
    Bradley-Terry Policy Optimization for Generative Preference Modeling
    Shengyu Feng, Yun He, Shuang Ma, Beibin Li, Yuanhao Xiong, Vincent Li, Karishma Mandyam, Julian Katz-Samuels, Shengjie Bi, Licheng Yu, Hejia Zhang, Karthik Abinav Sankararaman, Han Fang, Yiming Yang, and Manaal Faruqui
    arXiv preprint
    HTML
  5. ICLR
    FrontierCO: Real-World and Large-Scale Evaluation of Machine Learning Solvers for Combinatorial Optimization
    Shengyu Feng*, Weiwei Sun*, Shanda Li, Ameet Talwalker, and Yiming Yang
    International Conference on Learning Representations (ICLR), 2026.
    HTML Code
  6. AAAI
    CO-Bench: Benchmarking Language Model Agents in Algorithm Search for Combinatorial Optimization
    Weiwei Sun*, Shengyu Feng*, Shanda Li, and Yiming Yang
    Annual AAAI Conference on Artificial Intelligence (AAAI), 2026.
    HTML Code
  7. FGCS
    Alternative Mixed Integer Linear Programming Optimization for Joint Job Scheduling and Data Allocation in Grid Computing
    Shengyu Feng*, Jaehyung Kim*, Yiming Yang, Joseph Boudreau, Tasnuva Chowdhury, Adolfy Hoisie, Raees Khan, Ozgur O. Kilic, Scott Klasky, Tatiana Korchuganova, Paul Nilsson, Verena Ingrid Martinez Outschoorn, David K. Park, Norbert Podhorszki, Yihui Ren, Frederic Suter, Sairam Sri Vatsavai, Wei Yang, Shinjae Yoo, Tadashi Maeno, and Alexei Klimentov
    Future Generation and Computer Systems (FGCS).
    HTML
  8. ICML
    Regularized Langevin Dynamics for Combinatorial Optimization
    Shengyu Feng and Yiming Yang
    International Conference on Machine Learning (ICML), 2025.
    HTML Code
  9. ICLR
    Step-by-Step Reasoning for Math Problems via Twisted Sequential Monte Carlo
    Shengyu Feng, Xiang Kong, Shuang Ma, Aonan Zhang, Dong Yin, Chong Wang, Ruoming Pang, and Yiming Yang
    International Conference on Learning Representations (ICLR), 2025.
    HTML Code
  10. AAAI
    SORREL: Suboptimal-Demonstration-Guided Reinforcement Learning for Learning to Branch
    Shengyu Feng, Yiming Yang
    Annual AAAI Conference on Artificial Intelligence (AAAI), 2026.
    HTML Code Oral Presentation (top 6%)
  11. SDM
    Concept Discovery for Fast Adaptation
    Shengyu Feng and Hanghang Tong
    SIAM International Conference on Data Mining (SDM), 2023.
    HTML
  12. WACV
    Exploiting Long-Term Dependencies for Generating Dynamic Scene Graphs
    Shengyu Feng, Subarna Tripathi, Hesham Mostafa, Marcel Nassar, and Somdeb Majumdar.
    IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023.
    HTML Code
  13. TKDD
    ARIEL: Adversarial Graph Contrastive Learning
    Shengyu Feng, Baoyu Jing, Yada Zhu, and Hanghang Tong
    ACM Transactions on Knowledge Discovery from Data (TKDD).
    HTML Code
  14. CIKM
    X-GOAL: Multiplex Graph Prototypical Contrastive Learning
    Baoyu Jing, Shengyu Feng, Yuejia Xiang, Xi Chen, Yu Chen, and Hanghang Tong
    ACM International Conference on Information and Knowledge Management (CIKM), 2022.
    HTML
  15. WWW
    Adversarial Graph Contrastive Learning with Information Regularization
    Shengyu Feng, Baoyu Jing, Yada Zhu, and Hanghang Tong
    ACM Web Conference (WWW), 2022.
    HTML Code
  16. ICLR
    Batch Reinforcement Learning through Continuation Method
    Yijie Guo, Shengyu Feng, Nicolas Le Roux, Ed Chi, Honglak Lee, and Minmin Chen
    International Conference on Learning Representations (ICLR), 2021.
    HTML
  17. NeurIPS
    Memory Based Trajectory-conditioned Policies for Learning from Sparse Rewards
    Yijie Guo, Jongwook Choi, Marcin Moczulski, Shengyu Feng, Samy Bengio, Mohammad Norouzi, and Honglak Lee
    Neural Information Processing Systems (NeurIPS), 2020.
    HTML

Invited Talks

Regularized Langevin Dynamics for Combinatorial Optimization
2025 INFORMS Annual Meeting, oral presentation
Oct. 2025
Benchmarking LLM Agents in Algorithm Search
Massachusetts Institute of Technology (MIT), invited benchmark talk
April 2025

Powered by Jekyll and Minimal Light theme.