Bocheng Li  [GitHub] [Google Scholar]

Ph.D. Student in Computer Science

University of Science and Technology of China (USTC)

State Key Laboratory of Cognitive Intelligence

Email: bcli (at) mail.ustc.edu.cn

About Me

I am a fourth-year Ph.D. student in Computer Science at the University of Science and Technology of China (USTC), advised by Prof. Linli Xu. I received my B.Eng. in Computer Science from USTC in 2023.

My research centers on making diffusion and flow-matching generative models faster and more practical. I have worked on few-step sampling for language flows, unified diffusion language models, visual tokenization for autoregressive image generation, and flow-based action policies for robotics. More recently, I am exploring efficient training and inference for large-scale generative models.

Outside research, I spent a wonderful year with my friends at USTC Linux User Group. My favorite band at the moment is ChiliChill. 

Publications
(* = equal contribution)

Efficient Diffusion, Flow, and Hybrid Generation:

• Improving Few-Step Language Flows with Untied Self-Conditioning [arXiv]
  Bocheng Li and Linli Xu
  arXiv preprint 2608.22244
  A training-free sampler that corrects a train–inference mismatch in self-conditioned flow-matching language models, improving few-step generation without additional model evaluations. e.g., reducing GenPPL by 8.6× at 8 sampling steps on LangFlow. 

• Unifying Continuous and Discrete Text Diffusion with Non-simultaneous Diffusion Processes [arXiv] [ACL Anthology] [Code]
  Bocheng Li*, Zhujin Gao* and Linli Xu
  In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025), Oral Presentation
  A non-simultaneous continuous diffusion language model that combines fine-grained continuous noising with token-wise adaptive diffusion progress, providing a unified view of continuous and discrete diffusion language models.

• Few-shot Temporal Pruning Accelerates Diffusion Models for Text Generation  [ACL Anthology] [Code]
  Bocheng Li, Zhujin Gao, Yongxin Zhu, Kun Yin, Haoyu Cao, Deqiang Jiang and Linli Xu
  In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (COLING 2024)
  A training-free, few-shot sampling-schedule optimization method that removes redundant denoising steps across discrete and continuous diffusion language models, achieving up to 400× acceleration after less than one minute of optimization. 

• FastHybrid: Accelerating Hybrid Autoregressive Image Generation with Lookahead and Guided Decoding  [CVF OpenAccess]
  Zhengguo Jiang, Fang Zhang, Yongxiang Hua, Bocheng Li, Wentao Zhang and Linli Xu
  In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2026 (CVPR 2026)
  Accelerates hybrid autoregressive–diffusion image generation by combining lookahead decoding with guided diffusion sampling, improving the speed–quality trade-off of iterative visual generation. 

Visual Representation and Multimodal Learning:

• Addressing Representation Collapse in Vector Quantized Models with One Linear Layer [arXiv] [Code] [Post on kexue.fm] [@lucidrain's VQ Repo]
  Yongxin Zhu, Bocheng Li, Yifei Xin and Linli Xu
  In Proceedings of the International Conference on Computer Vision 2025 (ICCV 2025)
  Reparameterizes the VQ codebook as a learnable linear space so that all code vectors can be optimized jointly, mitigating representation collapse across image and audio models. Adopted by Kelix @ Kuaishou and many other research projects.

• Stabilize the Latent Space for Image Autoregressive Modeling: A Unified Perspective [arXiv] [Code]
  Yongxin Zhu, Bocheng Li, Hang Zhang, Xin Li, Linli Xu and Lidong Bing
  In Proceedings of the Thirty-eighth Annual Conference on Neural Information Processing Systems (NeurIPS 2024)
  Identifies latent-space stability as a key factor in image generative modeling and constructs discrete visual tokens from self-supervised features for autoregressive image generation and understanding. Among the earliest works of vision tokenizers for unified generative models.

Other super interesting things we've explored:

• Learning Native Continuation for Action Chunking Flow Policies [arXiv]
  Yufeng Liu, Hang Yu, Juntu Zhao, Bocheng Li, Di Zhang, Mingzhu Li, Wenxuan Wu, Yingdong Hu, Junyuan Xie, Junliang Guo, Dequan Wang and Yang Gao
  In Proceedings of Robotics: Science and Systems 2026 (RSS 2026)
  Training-time continuation method for action-chunked VLA policies that aligns training and inference dynamics to produce intrinsically smooth, temporally consistent actions, outperforming RTC/Train-time RTC(which is the one deployed on Pi0.6 @ Physical Intelligence) in both smoothness and efficiency.

• Input Domain Aware MoE: Decoupling Routing Decisions from Task Optimization in Mixture of Experts [arXiv]
  Yongxiang Hua, Haoyu Cao, Zhou Tao, Bocheng Li, Zihao Wu, Chaohu Liu and Linli Xu
  In Proceedings of the 33rd ACM International Conference on Multimedia (ACM MM 2025)
  A probabilistic routing framework that decouples expert assignment from task optimization, improving expert specialization, load balancing, and performance in vision-language models.

Experience

• Research Intern, Foundation Models @ Spirit AI Robotics, November 2025 - February 2026
  Worked on large-scale training and evaluation of foundation VLA policies.

• Research Intern, Tencent YouTu Research, March 2023 - October 2023
  Focused on Generative Models and MLLMs for VQA.

• President of USTC Linux User Group, May 2022 - May 2023
  USTCLUG is the 5-star student club at USTC in 2023, the highest honor bestowed upon student organizations.

Awards

• National Scholarship, October 2025

• USTC-Suzhou Industrial Park Scholarship, November 2024

• Excellence Award (Advanced Game Theory Algorithm, Top 10%), Tencent AIArena National Open Competition for Artificial Intelligence, December 2023

• USTC Outstanding Thesis Award (Bachelor Thesis; 83 out of 1860), June 2023

Teaching Assistant

• Basics of Artificial Intelligence, 2023, 2024 Spring (USTC)

• Computer Systems: A Programmer's Perspective[CS:APP], 2022 Spring (USTC)

Services

• Reviewer: ACL ARR, NeurIPS 2024/2025, ICLR 2025/2026, AAAI 2026, CVPR 2026