Quanquan Gu
Associate Professor
Develops nonconvex optimization and learning theory, and alignment and reasoning methods for large language models.
Quanquan Gu's lab at UCLA works on learning theory & optimization, trustworthy & safe ML and large language models. 91 papers since 2023, 73 at top venues, mostly ICML and ICLR. 60 alumni, most often to Amazon and ByteDance. Applications close .
At a glance
- papers, 2023–2026
- 24/ year
- at T1 venues
- 80%
- PhD students
- 0
- MS students
- 0
- alumni listed
- 60
Papers by yearfrom the lab homepage, 2023–2026
- 9 papers2026 · T1 9 · collecting
- 26 papers2025 · T1 21
- 25 papers2024 · T1 18
- 31 papers2023 · T1 25
T1 venuesOtherVenues: ICML, ICLR, TMLR, UAI, COLT
Research focuspaper titles that use each topic's words
- Learning theory & optimization33 papers
- Large language models24 papers
- nonconvex optimization2 papers
- nonconvex2 papers
- Trustworthy & safe ML0 papers
Paper list
From the lab homepage.
20269 papers
- ICLR (T1)Best-of-Majority: Minimax-Optimal Strategy for Pass@k Inference Scaling (opens in a new tab)
- ICLR (T1)Breaking the Total Variance Barrier: Sharp Sample Complexity for Linear Heteroscedastic Bandits with Fixed Action Set (opens in a new tab)
- ICLR (T1)Group Representational Position Encoding (opens in a new tab)
- ICLR (T1)On the Design of KL-Regularized Policy Gradient Algorithms for LLM Reasoning (opens in a new tab)
- ICLR (T1)Scalable Spatio-Temporal SE(3) Diffusion for Long-Horizon Protein Dynamics (opens in a new tab)
- ICLR (T1)SWINGARENA: Adversarial Programming Arena for Long-context GitHub Issue Solving (opens in a new tab)
- ICLR (T1)Towards a Sharp Analysis of Offline Policy Learning for f-Divergence-Regularized Contextual Bandits (opens in a new tab)
- ICLR (T1)Transformers Trained via Gradient Descent Can Provably Learn a Class of Teacher Models (opens in a new tab)
- ICLR (T1)Variance-Dependent Regret Lower Bounds for Contextual Bandits (opens in a new tab)
202526 papers
- ICML (T1)An All-Atom Generative Model for Designing Protein Complexes (opens in a new tab)
- ICML (T1)Beyond Bradley-Terry Models: A General Preference Model for Language Model Alignment (opens in a new tab)
- ICLR (T1)Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration (opens in a new tab)
- ICLR (T1)Convergence of Score-Based Discrete Diffusion Models: A Discrete-Time Analysis (opens in a new tab)
- ICLR (T1)CryoFM: A Flow-based Foundation Model for Cryo-EM Densities (opens in a new tab)
- TMLR (T1)Decomposed Direct Preference Optimization for Structure-Based Drug Design (opens in a new tab)
- ICML (T1)Designing Cyclic Peptides via Harmonic SDE with Atom-Bond Modeling (opens in a new tab)
- ACM Computing SurveysDomain Specialization as the Key to Make Large Language Models Disruptive: A Comprehensive Survey (opens in a new tab)
- ICLR (T1)DPLM-2: A Multimodal Diffusion Protein Language Model (opens in a new tab)
- ICML (T1)Elucidating the Design Space of Multimodal Protein Language Models (opens in a new tab)
- ICLR (T1)Energy-Weighted Flow Matching for Offline Reinforcement Learning (opens in a new tab)
- ICML (T1)Global Convergence and Rich Feature Learning in L-Layer Infinite-Width Neural Networks under muP-Parametrization (opens in a new tab)
202425 papers
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)A Nearly Optimal and Low-Switching Algorithm for Reinforcement Learning with General Function Approximation (opens in a new tab)
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Achieving Constant Regret in Linear Markov Decision Processes (opens in a new tab)
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Antigen-Specific Antibody Design via Direct Energy-based Preference Optimization (opens in a new tab)
- ICML (T1)Borda Regret Minimization for Generalized Linear Dueling Bandits
- ICLR (T1)DecompOpt: Controllable and Decomposed Diffusion Models for Structure-based Molecular Optimization (opens in a new tab)
- ICML (T1)Diffusion Language Models Are Versatile Protein Learners
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Enhancing Large Vision Language Models with Self-Training on Image Comprehension (opens in a new tab)
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Fast Sampling via Discrete Non-Markov Diffusion Models with Predetermined Transition Time (opens in a new tab)
- ICML (T1)Feel-Good Thompson Sampling for Contextual Dueling Bandits
- ICLR (T1)Horizon-free Reinforcement Learning in Adversarial Linear Mixture MDPs (opens in a new tab)
- ICLR (T1)How Many Pretraining Tasks Are Needed for In-Context Learning of Linear Regression? (opens in a new tab)
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Matching the Statistical Query Lower Bound for k -Sparse Parity Problems with Sign Stochastic Gradient Descent (opens in a new tab)
202331 papers
- ICLR (T1)A General Framework for Sample-Efficient Function Approximation in Reinforcement Learning (opens in a new tab)
- ACM/IMS Journal of Data ScienceBatched Neural Bandits (opens in a new tab)
- ICML (T1)Benign Overfitting for Two-layer ReLU Convolutional Neural Networks (opens in a new tab)
- UAI (T1)Benign Overfitting in Adversarially Robust Linear Classification (opens in a new tab)
- ICML (T1)Cooperative Multi-Agent Reinforcement Learning: Asynchronous Communication and Linear Function Approximation
- ICML (T1)Corruption-Robust Algorithms with Uncertainty Weighting for Nonlinear Contextual Bandits and Markov Decision Processes
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Corruption-Robust Offline Reinforcement Learning with General Function Approximation (opens in a new tab)
- ICML (T1)DecompDiff: Diffusion Models with Decomposed Priors for Structure-Based Drug Design
- UAI (T1)Efficient Privacy-Preserving Stochastic Nonconvex Optimization (opens in a new tab)
- ICML (T1)Finite-Sample Analysis of Learning High-Dimensional Single ReLU Neuron
- ICLR (T1)How Does Semi-supervised learing with Pseudo-labelers Work? A Case Study (opens in a new tab)
- in Proc. of Advances in Neural Information Processing Systems (NeurIPS)Implicit Bias of Gradient Descent for Two-layer ReLU and Leaky ReLU Networks on Nearly-orthogonal Data (opens in a new tab)
Where alumni went60 listed, 36% to industry
Industry 8Academia 14
Most often to
- Amazon
- ByteDance
- IBM
Also to
- Caltech
- Carnegie Mellon University
- Columbia University
Apply to UCLAFall 2027
Computer Science PhD
- Round
- Single round (Dec 15)
- Opens
- —
- GRE
- optional
- English
- See program page
- Fee
- See program page
Not yet confirmed for this intake.
Similar labsnearest by research summary
- CNSJTU
Chaoxiang He
- Trustworthy & safe ML
- USUSC
Sai Praneeth Karimireddy
- Learning theory & optimization
- CNZhejiang
Siliang Tang
- Multimodal & grounded language
- CNSJTU
Shuo Wang
- Trustworthy & safe ML
- CNSJTU
Yibo Yang
- Trustworthy & safe ML
- USNYU
Rachel Greenstadt
- Trustworthy & safe ML
Your notes
Save this lab to keep notes.
Data
Checked against the lab homepage.
- Collected
- Verified
- Papers
- The lab homepage
- Members
- Lab homepage
- Alumni
- Lab homepage
Tiers are venue classifications, not lab rankings.
UCLAcloses