The Curious Engineer

All blogs / AI / ML

The Berkeley Artificial Intelligence Research Blog

bair.berkeley.edu/blog

The BAIR Blog

AI / ML

’17’18’19’20’21’22’23’24’25’26
Posts
171 from 20 Jun 2017 to 29 Jul 2026
Posting
monthly (9 in the last year)
Kind
magazine
Feed
https://bair.berkeley.edu/blog/feed.xml

2026 8

  1. From CUDA to MLX: How K-Search Brings Decades of Kernel Expertise to Apple Silicon
  2. Teaching LLMs to Update Beliefs for Efficient Long-Horizon Interaction
  3. Intelligence is Free, Now What? Data Systems for, of, and by Agents
  4. 2026 BAIR Graduate Showcase
  5. Adaptive Parallel Reasoning: The Next Paradigm in Efficient Inference Scaling
  6. Gradient-based Planning for World Models at Longer Horizons
  7. Identifying Interactions at Scale for LLMs
  8. Information-Driven Design of Imaging Systems

2025 6

  1. RL without TD learning
  2. What exactly does word2vec learn?
  3. Whole-Body Conditioned Egocentric Video Prediction
  4. Defending against Prompt Injection with Structured Queries (StruQ) and Preference Optimization (SecAlign)
  5. Repurposing Protein Folding Models for Generation with Latent Diffusion
  6. Scaling Up Reinforcement Learning for Traffic Smoothing: A 100-AV Highway Deployment

2024 9

  1. Virtual Personas for Language Models via an Anthology of Backstories
  2. Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discrimination
  3. How to Evaluate Jailbreak Methods: A Case Study with the StrongREJECT Benchmark
  4. Dillon Bowen
  5. Are We Ready for Multi-Image Reasoning? Launching VHs: The Visual Haystacks Benchmark!
  6. TinyAgent: Function Calling at the Edge
  7. Modeling Extremely Large Images with xT
  8. 2024 BAIR Graduate Directory
  9. The Shift from Models to Compound AI Systems

2023 11

  1. Asymmetric Certified Robustness via Feature-Convex Neural Networks
  2. Ghostbuster: Detecting Text Ghostwritten by Large Language Models
  3. Goal Representations for Instruction Following
  4. Rethinking the Role of PPO in RLHF
  5. Training Diffusion Models with Reinforcement Learning
  6. On the Stepwise Nature of Self-Supervised Learning
  7. Generating 3D Molecular Conformers via Equivariant Coarse-Graining and Aggregated Attention
  8. GPT-4 + Stable-Diffusion = ?: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models
  9. Interactive Fleet Learning
  10. Koala: A Dialogue Model for Academic Research
  11. Fully Autonomous Real-World Reinforcement Learning with Applications to Mobile Manipulation

2022 12

  1. Keeping Learning-Based Control Safe by Regulating Distributional Shift
  2. Reverse engineering the NTK: towards first-principles architecture design
  3. Why do Policy Gradient Methods work so well in Cooperative MARL? Evidence from Policy Representation
  4. FIGS: Attaining XGBoost-level performance with the interpretability and speed of CART
  5. The Berkeley Crossword Solver
  6. Rethinking Human-in-the-Loop for Artificial Augmented Intelligence
  7. Designing Societally Beneficial Reinforcement Learning Systems
  8. Should I Use Offline RL or Imitation Learning?
  9. Offline RL Made Easier: No TD Learning, Advantage Reweighting, or Transformers
  10. Accelerating Ukraine Intelligence Analysis with Computer Vision on Synthetic Aperture Radar Imagery
  11. Unsupervised Skill Discovery with Contrastive Intrinsic Control
  12. imodels: leveraging the unreasonable effectiveness of rules

2021 25

  1. The Unsupervised Reinforcement Learning Benchmark
  2. Which Mutual Information Representation Learning Objectives are Sufficient for Control?
  3. Sequence Modeling Solutions for Reinforcement Learning Problems
  4. Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasets
  5. Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability
  6. RECON: Learning to Explore the Real World with a Ground Robot
  7. Why Generalization in RL is Difficult: Epistemic POMDPs and Implicit Partial Observability
  8. A First-Principles Theory of NeuralNetwork Generalization
  9. Designs from Data: Offline Black-Box Optimization via Conservative Training
  10. Making RL Tractable by Learning More Informative Reward Functions: Example-Based Control, Meta-Learning, and Normalized Maximum Likelihood
  11. Updates and Lessons from AI Forecasting
  12. PICO: Pragmatic Compression for Human-in-the-Loop Decision-Making
  13. Unsolved ML Safety Problems
  14. Distilling neural networks into wavelet models using interpretations
  15. What Can I Do Here? Learning New Skills by Imagining Visual Affordances
  16. Universal Weakly Supervised Segmentation by Pixel-to-Segment Contrastive Learning
  17. The Surprising Effectiveness of PPO in Cooperative Multi-Agent Games
  18. BASALT: A Benchmark for Learning from Human Feedback
  19. Learning What To Do by Simulating the Past
  20. An EPIC way to evaluate reward functions
  21. The Importance of Hyperparameter Optimization for Model-based Reinforcement Learning
  22. Pretrained Transformers as Universal Computation Engines
  23. Maximum Entropy RL (Provably) Solves Some Robust RL Problems
  24. Self-Supervised Policy Adaptation during Deployment
  25. The Successor Representation, $\gamma$-Models, and Infinite-Horizon Prediction

2020 26

  1. Does GPT-2 Know Your Phone Number?
  2. Offline Reinforcement Learning: How Conservative Algorithms Can Enable New Applications
  3. Learning State Abstractions for Long-Horizon Planning
  4. Goodhart’s Law, Diversity and a Series of Seemingly Unrelated Toy Problems
  5. Adapting on the Fly to Test Time Distribution Shift
  6. Reinforcement learning is supervised learning on optimized data
  7. Plan2Explore: Active Model-Building for Self-Supervised Visual Reinforcement Learning
  8. AWAC: Accelerating Online Reinforcement Learning with Offline Datasets
  9. AI Will Change the World.Who Will Change AI?We Will.
  10. Estimating the fatality rate is difficult but doable with better data
  11. Exploring Exploration: Comparing Children with RL Agents in Unified Environments
  12. Can RL From Pixels be as Efficient as RL From State?
  13. Decentralized Reinforcement Learning:Global Decision-Making viaLocal Economic Transactions
  14. D4RL: Building Better Benchmarks for Offline Reinforcement Learning
  15. Open Compound Domain Adaptation
  16. OmniTact: A Multi-Directional High-Resolution Touch Sensor
  17. Four Novel Approaches to Manipulating Fabric using Model-Free and Model-Based Deep Learning in Simulation
  18. Unsupervised Meta-Learning: Learning to Learn without Supervision
  19. The Ingredients of Real World Robotic Reinforcement Learning
  20. Making Decision Trees Accurate Again: Explaining What Explainable AI Did Not
  21. Robots Learning to Move like Animals
  22. Physically Realistic Attacks on Deep Reinforcement Learning
  23. Does On-Policy Data Collection Fix Errors in Off-Policy Reinforcement Learning?
  24. BADGR:The Berkeley Autonomous Driving Ground Robot
  25. Speeding Up Transformer Training and Inference By Increasing Model Size
  26. Large Scale Training at BAIR with Ray Tune

2019 30

  1. Emergent Behavior by Minimizing Chaos
  2. What is My Data Worth?
  3. Learning to Imitate Human Demonstrations via CycleGAN
  4. Model-Based Reinforcement Learning:Theory and Practice
  5. Data-Driven Deep Reinforcement Learning
  6. RoboNet: A Dataset for Large-Scale Multi-Robot Learning
  7. Prof. Anca Dragan Talks About Human-Robot Interaction for WIRED
  8. Can We Learn the Language of Proteins?
  9. Look then Listen: Pre-Learning Environment Representations for Data-Efficient Neural Instruction Following
  10. Collaborating with Humans Requires Understanding Them
  11. Functional RL with Keras and Tensorflow Eager
  12. Deep Dynamics Models for Dexterous Manipulation
  13. Sample Efficient Evolutionary Algorithm for Analog Circuit Design
  14. rlpyt: A Research Code Base for Deep Reinforcement Learning in PyTorch
  15. A Deep Learning Approach to Data Compression
  16. Evaluating and Testing Unintended Memorization in Neural Networks
  17. Learning to Learn with Probabilistic Task Embeddings
  18. 1000x Faster Data Augmentation
  19. Autonomous Vehicles for Social Good: Learning to Solve Congestion
  20. End-to-End Deep Reinforcement Learning without Reward Engineering
  21. Model-Based Reinforcement Learning from Pixels with Structured Latent Variable Models
  22. Large-Scale Long-Tailed Recognition in an Open World
  23. Robots that Learn to Adapt
  24. Robots that Learn to Use Improvised Tools
  25. CVPR 2019 Challenges on Domain Adaptation in Autonomous Driving
  26. Announcing the BAIR Open Research Commons
  27. Manipulation By Feel
  28. Assessing Generalization in Deep Reinforcement Learning
  29. Controlling False Discoveries in Large-Scale Experimentation: Challenges and Solutions
  30. Learning Preferences by Looking at the World

2018 22

  1. Soft Actor Critic—Deep Reinforcement Learning with Real-World Robots
  2. Scaling Multi-Agent Reinforcement Learning
  3. Building Gene Expression Atlases with Deep Generative Models for Single-cell Transcriptomics
  4. Visual Model-Based Reinforcement Learning as a Path towards Generalist Robots
  5. Physics-Based Learned Design: Teaching a Microscope How to Image
  6. AdaSearch: A Successive Elimination Approach to Adaptive Search
  7. Drilling Down on Depth Sensing and Deep Learning
  8. Learning Acrobatics by Watching YouTube
  9. Visual Reinforcement Learning with Imagined Goals
  10. Dexterous Manipulation with Reinforcement Learning: Efficient, General, and Low-Cost
  11. When Recurrent Models Don’t Need to be Recurrent
  12. One-Shot Imitation from Watching Videos
  13. BDD100K Blog Update
  14. BDD100K: A Large-scale Diverse Driving Video Database
  15. Delayed Impact of Fair Machine Learning
  16. TDM: From Model-Free to Model-Based Deep Reinforcement Learning
  17. Shared Autonomy via Deep Reinforcement Learning
  18. Towards a Virtual Stuntman
  19. Transfer Your Font Style with GANs
  20. Learning Robot Objectives from Physical Human Interaction
  21. Kernel Feature Selection via Conditional Covariance Minimization
  22. Ray: A Distributed System for AI

2017 22

  1. Physical Adversarial Examples Against Deep Neural Networks
  2. Reverse Curriculum Generation for Reinforcement Learning Agents
  3. Towards Intelligent Industrial Co-robots
  4. FaSTrack: Ensuring Safe Real-Time Navigation of Dynamic Systems
  5. Model-based Reinforcement Learning with Neural Network Dynamics
  6. The Emergence of a Fovea while Learning to Attend
  7. DART: Noise Injection for Robust Imitation Learning
  8. Learning Long Duration Sequential Task Structure From Demonstrations with Application in Surgical Robotics
  9. Learning Diverse Skills via Maximum Entropy Deep Reinforcement Learning
  10. Learning to Optimize with Reinforcement Learning
  11. Learning a Multi-View Stereo Machine
  12. How to Escape Saddle Points Efficiently
  13. High Quality 3D Object Reconstruction from a Single Color Image
  14. Cooperatively Learning Human Values
  15. Captioning Novel Objects in Images
  16. Minibatch Metropolis-Hastings
  17. Learning to Learn
  18. The Confluence of Geometry and Learning
  19. Constrained Policy Optimization
  20. Releasing the Dexterity Network (Dex-Net) 2.0 Dataset for Deep Grasping
  21. Introducing the BAIR Blog
  22. Learning to Reason with Neural Module Networks