Qinqing Zheng

About

• I am a ML researcher. Currently, I lead reinforcement learning for diffusion language models at Inception.
• Previously, I was on the Llama research team, specializing in post-training for reasoning. Within that effort, I spearheaded the development of the practical domain (non-math/non-coding) expert model. Before that, I worked at FAIR on a combination of RL, diffusion, and LLM reasoning.
• I received my Ph.D.from UChicago, advised by Prof. John Lafferty (a truly fortunate experience!). My doctoral research combined theory and practice, with a focus on exact solutions for nonconvex optimization problems.
  
  Earlier, I was a research scientist at Facebook, where I helped build the distributed training system supporting Ads ranking. I also worked as a postdoc researcher in Wharton Statistics, focusing on differential privacy and statistical inference.

My Notes

• On-Policy Distillation

Models

I contributed to

• Mercury 2

• Llama 4

Papers

• d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning
  Siyan Zhao*, Devaansh Gupta*, Qinqing Zheng†, Aditya Grover†
  NeurIPS 2025 (Spotlight)
  [paper] [code]

• Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning
  DiJia Su, Hanlin Zhu*, Yingchen Xu*, Jiantao Jiao, Yuandong Tian†, Qinqing Zheng†
  ICML 2025
  [paper]

• Online Intrinsic Rewards for Decision Making Agents from Large Language Model Feedback
  Qinqing Zheng*, Mikael Henaff*, Amy Zhang, Aditya Grover, Brandon Amos
  RLC 2025
  [paper] [code]

• Dualformer: Controllable Fast and Slow Thinking by Learning with Randomized Reasoning Traces
  DiJia Su, Sainaa Sukhbaatar, Michael Rabbat, Yuandong Tian, Qinqing Zheng
  ICLR 2025
  [paper] [code]

• Beyond A*: Better Planning with Transformers via Search Dynamics Bootstrapping
  Lucas Lehnert, Sainaa Sukhbaatar, DiJia Su, Qinqing Zheng, Paul Mcvay, Michael Rabbat, Yuandong Tian
  COLM 2024
  [paper] [code]

• Diffusion World Model: Future Modeling Beyond Step-by-Step Rollout for Offline Reinforcement Learning
  Zihan Ding, Amy Zhang, Yuandong Tian, Qinqing Zheng
  ICLR 2024 Generative Models for Decision Making Workshop
  [paper]

• Guided Flows for Generative Modeling and Decision Making
  Qinqing Zheng, Matt Le, Neta Shaul, Yaron Lipman, Aditya Grover, Ricky T. Q. Chen
  [paper]

• Dual RL: Unification and New Methods for Reinforcement and Imitation Learning
  Harshit Sikchi, Qinqing Zheng, Amy Zhang, Scott Niekum
  ICLR 2024 (Spotlight)
  [paper] [code]

• Semi-Supervised Offline Reinforcement Learning with Action-Free Trajectories
  Qinqing Zheng, Mikael Henaff, Brandon Amos, Aditya Grover
  ICML 2023
  [paper] [code]

• ConserWeightive Behavioral Cloning for Reliable Offline Reinforcement Learning
  Tung Nguyen, Qinqing Zheng, Aditya Grover
  NeurIPS 2022 Foundation Models for Decision Making Workshop
  [paper] [code]

• Latent State Marginalization as a Low-cost Approach for Improving Exploration
  Dinghuai Zhang, Aaron Courville, Yoshua Bengio, Qinqing Zheng, Amy Zhang, Ricky T. Q. Chen
  ICLR 2023
  [paper]

• Online Decision Transformer
  Qinqing Zheng, Amy Zhang, Aditya Grover
  ICML 2022 (Long Oral Presentation)
  [paper] [code] [poster]

• Near-Optimal Confidence Sequences for Bounded Random Variables
  Arun Kumar Kuchibhotla*, Qinqing Zheng* (*Equal contribution)
  ICML 2021 (Spotlight)
  [paper] [code]

• A Theorem of the Alternative for Personalized Federated Learning
  Shuxiao Chen, Qinqing Zheng, Qi Long, Weijie Su
  Submitted.
  [paper]

• Federated \(f\)-Differential Privacy
  Qinqing Zheng, Shuxiao Chen, Qi Long, Weijie Su
  AISTATS 2021
  [paper] [code]

• Sharp Composition Bounds for Gaussian Differential Privacy via Edgeworth Expansion
  Qinqing Zheng, Jinshuo Dong, Qi Long, Weijie Su
  ICML 2020
  [paper] [code]

• ShadowSync: Performing Synchronization in the Background for Highly Scalable Distributed Training
  Qinqing Zheng, Bor-Yiing Su, Jiyan Yang, Alisson Azzolini, Qiang Wu, Ou Jin, Shri Karandikar, Hagay Lupesko, Liang Xiong, Eric Zhou
  [paper]

• Convergence Analysis for Rectangular Matrix Completion Using Burer-Monteiro Factorization and Gradient Descent
  Qinqing Zheng, John Lafferty
  [paper]

• A Convergent Gradient Descent Algorithm for Rank Minimization and Semidefinite Programming from Random Linear Measurements
  Qinqing Zheng, John Lafferty
  NeurIPS 2015
  [paper] [poster]

• Interpolating Convex and Non-Convex Tensor Decompositions via the Subspace Norm
  Qinqing Zheng , Ryota Tomioka
  NeurIPS 2015
  [paper] [code] [poster]

Review

• Conference: ICML, NeurIPS, AISTATS, ICLR
• Journal: Journal of Machine Learning Research (Editorial Board Reviewer), IEEE Transactions on Signal Processing, Annals of Statistics