Jason Lee
Associate Professor
Studies the theory of deep learning, representation learning and reinforcement learning.
Jason Lee's lab at UC Berkeley works on reinforcement learning and learning theory & optimization. 76 papers since 2023, 75 at top venues, mostly ICLR and NeurIPS. 12 alumni, most often to Meta. Applications close .
At a glance
- papers, 2023–2026
- 20/ year
- at T1 venues
- 99%
- PhD students
- 0
- MS students
- 0
- alumni listed
- 12
Papers by yearfrom the lab homepage, 2023–2026
- 9 papers2026 · T1 9 · collecting
- 26 papers2025 · T1 25
- 21 papers2024 · T1 21
- 20 papers2023 · T1 20
T1 venuesOtherVenues: ICLR, NeurIPS, ICML, COLT, AISTATS
Research focuspaper titles that use each topic's words
- Learning theory & optimization36 papers
- Reinforcement learning34 papers
- representation learning0 papers
- representation0 papers
- computer sciences0 papers
Paper list
From the lab homepage.
20269 papers
- ICML (T1)Coverage Improvement and Fast Convergence of On-policy Preference Learning (opens in a new tab)
- ICLR (T1)Improved high-dimensional estimation with Langevin dynamics and stochastic weight averaging (opens in a new tab)
- ICLR (T1)Neural Networks Learn Generic Multi-Index Models Near Information-Theoretic Limit (opens in a new tab)
- COLT (T1)On the Statistical Query Complexity of Learning Semiautomata: a Random Walk Approach (opens in a new tab)
- COLT (T1)Provable Learning of Random Hierarchy Models and Hierarchical Shallow-to-Deep Chaining (opens in a new tab)
- ICLR (T1)Quantitative Bounds for Length Generalization in Transformers (opens in a new tab)
- COLT (T1)Risk Comparisons in Linear Regression: Implicit Regularization Dominates Explicit Regularization (opens in a new tab)
- ICML (T1)Statistical Learning Theory in Lean 4: Empirical Processes from Scratch (opens in a new tab)
- ICML (T1)The Power of Power Law: Asymmetry Enables Compositional Reasoning (opens in a new tab)
202526 papers
- NeurIPS (T1)Accelerating RL for LLM Reasoning with Optimal Advantage Regression (opens in a new tab)
- COLT (T1)Anytime Acceleration of Gradient Descent (opens in a new tab)
- ICLR (T1)Correcting the Mythos of KL-Regularization: Direct Alignment without Overoptimization via Chi-Squared Preference Optimization (opens in a new tab)
- NeurIPS (T1)DAG-Math: Graph-Guided Mathematical Reasoning in LLMs (opens in a new tab)
- NeurIPS (T1)Deployment Efficient Reward-Free Exploration with Linear Function Approximation (opens in a new tab)
- ICML (T1)Discrepancies are Virtue: Weak-to-Strong Generalization through Lens of Intrinsic Dimension (opens in a new tab)
- NeurIPS (T1)Emergence and scaling laws in SGD learning of shallow neural networks (opens in a new tab)
- ICLR (T1)Exploiting Structure in Offline Multi-Agent RL: The Benefits of Low Interaction Rank (opens in a new tab)
- AISTATS (T1)How Well Can Transformers Emulate In-context Newton's Method? (opens in a new tab)
- COLT (T1)Learning Compositional Functions with Transformers from Easy-to-Hard Data (opens in a new tab)
- ICLR (T1)Learning Hierarchical Polynomials of Multiple Nonlinear Features with Three-Layer Networks (opens in a new tab)
- NeurIPS (T1)Learning Orthogonal Multi-Index Models: A Fine-Grained Information Exponent Analysis (opens in a new tab)
202421 papers
- ICML (T1)An Information-Theoretic Analysis of In-Context Learning (opens in a new tab)
- NeurIPS (T1)BitDelta: Your Fine-Tune May Only Be Worth One Bit (opens in a new tab)
- COLT (T1)Computational-Statistical Gaps in Gaussian Single-Index Models (opens in a new tab)
- ICLR (T1)Dichotomy of Early and Late Phase Implicit Biases Can Provably Induce Grokking (opens in a new tab)
- ICLR (T1)Horizon-Free Regret for Linear Markov Decision Processes (opens in a new tab)
- ICML (T1)How Transformers Learn Causal Structure with Gradient Descent (opens in a new tab)
- NeurIPS (T1)Learning and Transferring Sparse Contextual Bigrams with Linear Transformers (opens in a new tab)
- ICLR (T1)Learning Hierarchical Polynomials with Three-Layer Neural Networks (opens in a new tab)
- ICML (T1)LoRA Training in the NTK Regime has No Spurious Local Minima (opens in a new tab)
- ICML (T1)Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads (opens in a new tab)
- NeurIPS (T1)Neural network learns low-dimensional polynomials with SGD near the information-theoretic limit (opens in a new tab)
- ICLR (T1)Provable Offline Reinforcement Learning with Human Feedback (opens in a new tab)
202320 papers
- ICLR (T1)Can We Find Nash Equilibria at a Linear Rate in Markov Games? (opens in a new tab)
- ICML (T1)Computationally Efficient PAC RL in POMDPs with Latent Determinism and Conditional Embeddings (opens in a new tab)
- ICLR (T1)Decentralized Optimistic Hyperpolicy Mirror Descent: Provably No-Regret Learning in Markov Games (opens in a new tab)
- ICML (T1)Efficient displacement convex optimization with particle gradient descent
- NeurIPS (T1)Fine-Tuning Language Models with Just Forward Passes (opens in a new tab)
- NeurIPS (T1)Implicit Bias of Gradient Descent for Logistic Regression at the Edge of Stability (opens in a new tab)
- ICML (T1)Local Optimization Achieves Global Optimality in Multi-Agent Reinforcement Learning (opens in a new tab)
- ICML (T1)Looped Transformers as Programmable Computers (opens in a new tab)
- NeurIPS (T1)Offline Minimax Soft-Q-learning Under Realizability and Partial Coverage (opens in a new tab)
- AISTATS (T1)Optimal Sample Complexity Bounds for Non-convex Optimization under Kurdyka-Lojasiewicz Condition
- ICLR (T1)PAC Reinforcement Learning for Predictive State Representations (opens in a new tab)
- NeurIPS (T1)Provable Guarantees for Nonlinear Feature Learning in Three-Layer Neural Networks (opens in a new tab)
Where alumni went12 listed, 14% to industry
Industry 1Academia 6
Most often to
- Meta
Also to
- Ben-Gurion University
- Harvard University
- HKUST
Apply to UC BerkeleyFall 2027
UC Berkeley EECS Computer Science PhD
- Round
- Single round
- Opens
- —
- GRE
- not required
- English
- TOEFL/IELTS (non-English-country degrees)
- Fee
- See program page
Confirmed on the program page.
Similar labsnearest by research summary
- USUSC
Paria Rashidinejad
- Reinforcement learning
- USMIT
Ankur Moitra
- Trustworthy & safe ML
- USUW
Natasha Jaques
- Reinforcement learning
- CNPeking
Ruosong Wang
- Learning theory & optimization
- DETUM
Gjergji Kasneci
- Trustworthy & safe ML
- USCMU
Andrej Risteski
- Learning theory & optimization
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.
UC Berkeleycloses