Somayeh Sojoudi
Associate Professor in Residence
Studies optimization and robustness of AI systems, with applications to power systems.
Somayeh Sojoudi's lab at UC Berkeley works on trustworthy & safe ML and learning theory & optimization. 45 papers since 2023, 26 at top venues, mostly CDC and NeurIPS. 8 PhD students. 13 alumni. Applications close .
At a glance
- papers, 2023–2026
- 12/ year
- at T1 venues
- 58%
- PhD students
- 8
- MS students
- 0
- alumni listed
- 13
Papers by yearfrom the lab homepage, 2023–2026
- 8 papers2026 · T1 6 · collecting
- 10 papers2025 · T1 6
- 10 papers2024 · T1 6
- 17 papers2023 · T1 8
T1 venuesOtherVenues: CDC, NeurIPS, ICML, TMLR, AISTATS
Research focuspaper titles that use each topic's words
- Learning theory & optimization13 papers
- power4 papers
- Trustworthy & safe ML1 paper
- power systems0 papers
- studies optimization0 papers
Paper list
From the lab homepage.
20268 papers
- NeurIPS (T1)Agents’ Last Exam (opens in a new tab)
- ICML (T1)Breaking the Reversal Curse in Autoregressive Language Models via Identity Bridge (opens in a new tab)
- ICML (T1)Falsifying Sparse Autoencoder Reasoning Features in Language Models (opens in a new tab)
- ICML (T1)Learning for Diffusion Models: A Statistical Theory under Semi-Supervised Learning (opens in a new tab)
- ACCMulti-Agent Guided Policy Search for Non-Cooperative Dynamic Games (opens in a new tab)
- NeurIPS (T1)Provably Learn Graph Search: Training Dynamics and the Exponential Power of Depth (opens in a new tab)
- ICLR (T1)Reinforcement Learning via Value Gradient Flow (opens in a new tab)
- CDCStrategic Arbitrage in Electricity Markets: A Multi-Agent Policy Optimization Framework (opens in a new tab)
202510 papers
- RA-LCertifiable Reachability Learning Using a New Lipschitz Continuous Value Function (opens in a new tab)
- CDCCoordinating Distributed Energy Resources with Nodal Pricing in Distribution Networks: a Game-Theoretic Approach (opens in a new tab)
- IEEE Transactions on Power SystemsDistributionally Robust Joint Chance-Constrained Optimal Power Flow Using Relative Entropy (opens in a new tab)
- TMLR (T1)DRAGON: Distributional Rewards Optimize Diffusion Generative Models (opens in a new tab)
- TAC (T1)Evolutionary games on infinite strategy sets: Convergence to Nash equilibria via dissipativity (opens in a new tab)
- NeurIPS (T1)Generalization or Hallucination? Understanding Out-of-Context Reasoning in Transformers (opens in a new tab)
- IEEE Transactions on Control Systems TechnologyIntent Demonstration in General-Sum Dynamic Games via Iterative Linear-Quadratic Approximations (opens in a new tab)
- NeurIPS (T1)OVERT: A Benchmark for Over-Refusal Evaluation on Text-to-Image Models (opens in a new tab)
- NeurIPS (T1)Revising and Falsifying Sparse Autoencoder Feature Explanations (opens in a new tab)
- JMLR (T1)Towards optimal branching of linear and semidefinite relaxations for neural network robustness certification (opens in a new tab)
202410 papers
- AISTATS (T1)Absence of spurious solutions far from ground truth: A low-rank analysis with high-order losses (opens in a new tab)
- CDCApproximately Gaussian Replicator Flows: Nonconvex Optimization as a Nash-Convergent Evolutionary Game (opens in a new tab)
- Interspeech (T1)ConsistencyTTA: Accelerating Diffusion-Based Text-to-Audio Generation with Consistency Distillation (opens in a new tab)
- TMLR (T1)MixedNUTS: Training-Free Accuracy-Robustness Balance via Nonlinearly Mixed Classifiers (opens in a new tab)
- L4DCMixing Classifiers to Alleviate the Accuracy-Robustness Trade-Off (opens in a new tab)
- ICML (T1)Pausing Policy Learning in Non-stationary Reinforcement Learning (opens in a new tab)
- EMNLP (T1)Ranking Manipulation for Conversational Search Engines (opens in a new tab)
- L4DCSoft Convex Quantization: Revisiting Vector Quantization with Convex Optimization (opens in a new tab)
- SIOPTThe computation of approximate feedback Stackelberg equilibria in multi-player nonlinear constrained dynamic games (opens in a new tab)
- ICML (T1)Transport of Algebraic Structure to Latent Embeddings (opens in a new tab)
202317 papers
- Mathematical ProgrammingA New Complexity Metric for Nonconvex Rank-one Generalized Matrix Completion (opens in a new tab)
- NeurIPS (T1)Algorithmic Regularization in Tensor Optimization: Towards a Lifted Approach in Matrix Sensing (opens in a new tab)
- NeurIPS (T1)Asymmetric Certified Robustness via Feature-Convex Neural Networks (opens in a new tab)
- AAMAS (T1)Cost Inference for Feedback Dynamic Games from Noisy Partial State Observations and Incomplete Trajectories (opens in a new tab)
- HICSS-56Distributed Power System State Estimation Using Graph Convolutional Neural Networks (opens in a new tab)
- CDCDistributionally Robust Optimization for Nonconvex QCQPs with Stochastic Constraints (opens in a new tab)
- INFORMS Journal on OptimizationGeometric Analysis of Noisy Low-rank Matrix Recovery in the Exact Parameterized and the Overparameterized Regimes (opens in a new tab)
- CDCInitial State Interventions for Deconfounded Imitation Learning (opens in a new tab)
- CDCLeveraging the Physics of AC Power Flow in Support Vector Regression to Identify Power System Topology (opens in a new tab)
- CDCMeta-Learning Parameterized First-Order Optimizers using Differentiable Convex Optimization (opens in a new tab)
- AISTATS (T1)Noisy Low-rank Matrix Optimization: Geometry of Local Minima and Convergence Rate (opens in a new tab)
- ICML (T1)Over-parametrization via Lifting for Low-rank Matrix Sensing: Conversion of Spurious Solutions to Strict Saddle Points (opens in a new tab)
Where alumni went13 listed
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
- USUCLA
Yuan Tian
- Trustworthy & safe ML
- USNYU
Shengjie Wang
- Learning theory & optimization
- USUC Berkeley
Song Mei
- Learning theory & optimization
- USUT Austin
Aryan Mokhtari
- Learning theory & optimization
- USNYU
Mehryar Mohri
- Learning theory & optimization
- USMichigan
Clayton Scott
- 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