ADVERTISE WITH USReach Embedded Science readers →
UniStack EECS
Artificial Intelligence Laboratory, Stanford
- Language Model Pretraining Language models (LMs), like BERT 1 and the GPT series 2, achieve remarkable performance on many natural language processing (NLP) tasks. They are now the foundation of today’s NLP systems. 3 These models serve important roles in products and tools that we use every day, such as search engines like Google 4 […]
- The 60th Annual Meeting of the Association for Computational Linguistics (ACL) 2022 is taking place May 22nd – May 27th. We’re excited to share all the work from SAIL that’s being presented, and you’ll find links to papers, videos and blogs below. Feel free to reach out to the contact authors directly to learn more […]
- The International Conference on Learning Representations (ICLR) 2022 is being hosted virtually from April 25th – April 29th. We’re excited to share all the work from SAIL that’s being presented, and you’ll find links to papers, videos and blogs below. Feel free to reach out to the contact authors directly to learn more about the […]
- Discovering systematic errors with cross-modal embeddings In this blog post, we introduce Domino, a new approach for discovering systematic errors made by machine learning models. We also discuss a framework for quantitatively evaluating methods like Domino. Links: 📄 Paper (ICLR 2022) 🌍 Longer Walkthrough 💻 GitHub 📘 Docs 📒 Google Colab Machine learning models that […]
- [Summary] tl;dr: A tremendous amount of effort has been poured into training AI algorithms to competitively play games that computers have traditionally had trouble with, such as the retro games published by Atari, Go, DotA, and StarCraft II. The practical machine learning knowledge accumulated in developing these algorithms has paved the way for people to […]
- Deep models require a lot of training examples, but labeled data is difficult to obtain. This motivates an important line of research on leveraging unlabeled data, which is often more readily available. For example, large quantities of unlabeled image data can be obtained by crawling the web, whereas labeled datasets such as ImageNet require expensive […]
- The 36th AAAI Conference on Artificial Intelligence (AAAI 2022) is being hosted virtually from February 22th – March 1st. We’re excited to share all the work from SAIL that’s being presented, and you’ll find links to papers, videos and blogs below. Feel free to reach out to the contact authors directly to learn more about […]
- Introduction In 2019, Stanford entered the Alexa Prize Socialbot Grand Challenge 3 for the first time, with its bot Chirpy Cardinal, which went on to win 2nd place in the competition. In our previous post, we discussed the technical structure of our socialbot and how developers can use our open-source code to develop their own. […]
- This work was conducted as part of SAIL and CRFM. Deep learning has enabled improvements in the capabilities of robots on a range of problems such as grasping 1 and locomotion 2 in recent years. However, building the quintessential home robot that can perform a range of interactive tasks, from cooking to cleaning, in novel […]
- TL;DR Want something better than \(k\)-means? Our state-of-the-art \(k\)-medoids algorithm from NeurIPS, BanditPAM, is now publicly available! \(\texttt{pip install banditpam}\) and you're good to go! Like the \(k\)-means problem, the \(k\)-medoids problem is a clustering problem in which our objective is to partition a dataset into disjoint subsets. In \(k\)-medoids, however, we require that the […]
EECS, Berkeley
- Figure 1: CUDA-to-MLX optimization translation map. CUDA optimization knowledge can be translated into architecture-native MLX strategies rather than copied instruction-for-instruction. We face a new epoch in computing. Hardware is changing rapidly — not just faster GPUs, but a growing range of chips from different vendors, each with its own architecture and often tailored to specific […]
- .abbel-fig { display: block; text-align: center; margin: 2.4em 0; line-height: 1.4; max-width: 100%; } .abbel-fig img { display: block; margin: 0.65em auto 0; height: auto; max-width: 100%; } /* Image sizes; captions use a narrower measure below */ .abbel-fig–wide img { width: 100%; max-width: 100%; } .abbel-fig–wide-90 img { width: 100%; max-width: 90%; } .abbel-fig–wide-lg […]
- … government of the people, by the people, for the people … — Abraham Lincoln, Gettysburg Address (1863) The cost of AI is dropping rapidly. GPT-4-class capabilities cost roughly $30 per million tokens in early 2023; today the same runs under $1, and some providers are pushing costs below $0.10. Across benchmarks, inference prices have […]
- Congratulations to the Berkeley Artificial Intelligence Research (BAIR) Lab class of 2026! This year, BAIR celebrates another remarkable group of Ph.D. graduates whose curiosity, creativity, and perseverance have pushed the frontiers of artificial intelligence and machine learning. Their work spans the breadth of modern AI — robotics and embodied intelligence, large language models and reasoning, […]
- .apr-fig { text-align: center; margin: 1.35em 0; line-height: 1.4; } .apr-fig–wide img { display: inline-block; width: 100%; max-width: 100%; height: auto; vertical-align: middle; } .apr-fig–wide-0-8 { max-width: 80%; margin-left: auto; margin-right: auto; } .apr-fig–tall img { display: inline-block; max-height: 300px; width: auto; max-width: 100%; height: auto; object-fit: contain; vertical-align: middle; } .apr-fig–tall-1-2x img { display: […]
- .grasp-results-table table { font-size: 0.875rem; line-height: 1.35; width: 100%; } .grasp-results-table th, .grasp-results-table td { padding: 0.35rem 0.5rem; } /* Consistent whitespace between major sections (this post is long and hr-heavy) */ article.post-content h2 { margin-top: 2.75rem; margin-bottom: 0.75rem; } article.post-content h2:first-of-type { margin-top: 2.25rem; } article.post-content h3 { margin-top: 1.65rem; margin-bottom: 0.5rem; } article.post-content […]
- Understanding the behavior of complex machine learning systems, particularly Large Language Models (LLMs), is a critical challenge in modern artificial intelligence. Interpretability research aims to make the decision-making process more transparent to model builders and impacted humans, a step toward safer and more trustworthy AI. To gain a comprehensive understanding, we can analyze these systems […]
- An encoder (optical system) maps objects to noiseless images, which noise corrupts into measurements. Our information estimator uses only these noisy measurements and a noise model to quantify how well measurements distinguish objects. Many imaging systems produce measurements that humans never see or cannot interpret directly. Your smartphone processes raw sensor data through algorithms before […]
- In this post, I’ll introduce a reinforcement learning (RL) algorithm based on an “alternative” paradigm: divide and conquer. Unlike traditional methods, this algorithm is not based on temporal difference (TD) learning (which has scalability challenges), and scales well to long-horizon tasks. We can do Reinforcement Learning (RL) based on divide and conquer, instead of temporal […]
- What exactly does word2vec learn, and how? Answering this question amounts to understanding representation learning in a minimal yet interesting language modeling task. Despite the fact that word2vec is a well-known precursor to modern language models, for many years, researchers lacked a quantitative and predictive theory describing its learning process. In our new paper, we […]
Stars, Toronto
- by Jonathan KellyWe were delighted to receive the Best Paper Award at the 2026 IEEE International Conference on Multisensor Fusion and Integration (MFI 2026), held in Pilsen, Czech Republic. The paper, “Lost in Time? Continuous Symmetry and Identifiability in Aided Inertial Navigation with Unknown Measurement Delays,” by Jonathan Kelly, Phone Thiha Kyaw, and Matthew Giamou. Thanks to […]
- by Jonathan KellyIn collaboration with researchers from KTH Royal Institute of Technology and MIT, we’re organizing the RSS 2026 Workshop: “The Geometry of Motion: Physics-Informed Structures for Learning and Control,” happening on Friday, July 17 at UTS in Sydney, Australia! The workshop will revisit robot motion from a geometric and physics-informed perspective. We aim to explore how […]
- by Jonathan KellyWe’re delighted to welcome our 2026 cohort of summer research students! Three students from Engineering Science at U of T, Ben Wei, Alison Lei, and Oliver Petrovic (left to right in the image below), will be working on various lab projects, including high-performance planning and dexterous manipulation. Ben is entering his second year of EngSci, […]
- by Jonathan KellyCongratulations to recent graduate Dr. Oliver Limoyo, whose research work on PhotoBot was featured in the November 23, 2024 edition of IEEE Spectrum Robotics News! You can read more about PhotoBot in this short article.
- by Jonathan KellyProf. Kelly’s Canada Research Chair (Tier II) in Collaborative Robotics has been renewed for a second five-year term. The Laboratory is grateful to NSERC and the Government of Canada for supporting this important program and our research work! A short press release is available from University of Toronto Engineering News.
- by Jonathan KellyMembers of the STARS Laboratory will be at IROS 2024 in Abu Dhabi! We will present two papers on work led by Oliver Limoyo: O. Limoyo, J. Li, D. Rivkin, J. Kelly, and G. Dudek, “Photobot: Reference-guided interactive photog- raphy via natural language,” in Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems […]
- by Jonathan KellyOur own Prof. Kelly received the IEEE Robotics & Automation Magazine’s Outstanding Associated Editor Award for 2024. The award was presented at ICRA 2024 in Yokohama, Japan. Congratulations to all the awardees!
- by Jonathan KellyWe’re celebrating 10 years of research in the STARS Laboratory in 2023-2024! The lab was founded and began operating in June of 2013. We’d like to thank all of the lab members and research partners for making the last decade an exiting one! Robotics continues to grow rapidly as a field, and we’re looking forward […]
- by Jonathan KellyThanks to Phys.org for running a feature all about the recent preprint article by Olivier Lamarre (STARS) and Shantanu Malhotra (NASA JPL)! The article describes how to define recovery policies to ensure that a solar-powered rover remains safe (i.e., limits the risk of complete battery depletion) while travelling through permanently shadowed regions (at, e.g., the […]
- by Jonathan KellyThe University of Toronto Faculty of Applied Science and Engineering has released their annual list of Grads to Watch (2023), and we’re delighted that STARS Laboratory alumnus Flip Marić is part of the cohort! Filip’s PhD work focused on geometric methods for robotic manipulation; he was a Joint Educational Placement (JEP) student, co-supervised by Ivan […]
ZitNik, Harvard
- by Marinka ZitnikFive papers accepted to NeurIPS 2026: (1) AutoScientists with self-organizing agent teams for long-running scientific experimentation, (2) Prompting decision transformers for zero-shot reach-avoid policies, (3) How post-training shapes biological reasoning models, (4) STRAND with sequence-conditioned transport for single-cell perturbations, and (5) StereoPep on molecular models for stereochemistry modeling.
- by Marinka ZitnikBiomedical world models are AI systems that simulate how biological systems evolve under interventions. [Cell paper]
- by Marinka ZitnikIn this piece, we reflect on 25 years of cancer research, including agentic AI and human-AI co-science, and what it will take for them to work in cancer research.
- by Marinka ZitnikThis moonshot project will turbocharge our AI research to understand human brain development and provide insights into autism, all in collaboration with Arlotta and Sestan Labs.
- by Marinka ZitnikNew paper on closing the loop from hypothesis to experiment, and on long-horizon reasoning in autonomous biomedical discovery.
- by Marinka ZitnikCongratulations to Michelle, Zhenglun, and Xiang, who are completing their postdoctoral training and starting tenure-track Assistant Professor positions.
- by Marinka ZitnikCongratulations to Amy, Michelle, Pengwei, and Richard, who are moving on to start their PhDs at MIT, Harvard, Columbia, and Tsinghua University.
- by Marinka ZitnikWe spend a lot of time developing LLMs and agents that can answer the questions we ask them. But the reverse direction gets much less attention: when the prompt does not contain enough information, can a model recognize what is missing and ask for it? [New paper] [Project website]
- by Marinka ZitnikAttend and submit your papers to our AI Scientists Workshop at NeurIPS 2026.
- by Marinka ZitnikMedea is an AI agent that nominates therapeutic targets across cell type contexts, predicts synthetic lethality in cancer cell lines, and forecasts immunotherapy response from multimodal patient profiles. To test whether Medea can identify the genetic vulnerabilities that DNA-damaging treatments exploit, we use it to scan 238,046 gene combinations for synthetic lethality. [Project website]
EECS, MIT
- by Jane Halpern | Department of Electrical Engineering and Computer ScienceIn the MIT class 6.C31 (Hack Yourself) students apply computer science strategies to personal growth goals.
- by Adam Zewe | MIT NewsBy rethinking how large cloud computing systems operate, Associate Professor Christina Delimitrou seeks to make data centers more energy efficient.
- Margaret Hamilton, computing pioneer who led software development for the Apollo program, dies at 90by Janine Liberty | Department of Aeronautics and AstronauticsOver two decades at MIT, Hamilton forged new paths in computer science and engineering. Her groundbreaking work included leading the team that programmed the first computers that took humans to the moon.
- by Michaela Jarvis | MIT Laboratory for Information and Decision SystemsAssociate Professor Cathy Wu uses reinforcement learning to help map out improvements to transportation and other multifaceted systems.
- by Alex Shipps | MIT CSAIL“InstructMesh” can generate designs for everyday objects that are easy to edit and fabricate for both experts and newcomers to 3D modeling.
- by Adam Zewe | MIT NewsAble to defeat top-ranked human players and more efficient than other models, the new system could help decision-makers in military maneuvers or business negotiations.
- by Adam Zewe | MIT NewsIn a study focusing on hiring decisions, MIT researchers found the use of a single algorithm by many firms could benefit job seekers in certain situations.
- by Anne Trafton | MIT NewsMIT engineers have found a way to stabilize the lipid nanoparticles used to deliver RNA vaccines, which could allow the vaccines to be more widely distributed.
- by David Chandler | Office of the Vice President for Energy and ClimateThe Climate Grand Challenges “New World of Weather” project shows how modeling, planning tools, and infrastructure analysis are translating research into real-world resilience.
- by Jordan Silva | School of EngineeringFaculty members and researchers were honored in recognition of their scholarship, service, and overall excellence.
Science & Technology, Yale
- Yale researchers show that networks of flowing water and other fluids can store their own memories — similar to how a computer chip does — an innovation that could steer robots and provide insight about some simple organisms.
- Yale researchers help prove the utility of an innovative radio telescope in understanding the expanding universe and offer thoughts on how to best use AI in science. Meanwhile, faculty members in physics and astronomy are honored for breakthroughs across an array of fields.
- Yale Engineering’s Theodore Kim developed hair simulation math that Pixar adapted for Blaze’s tightly coiled hair in “Toy Story 5.”
- The National Academies of Sciences, Engineering, and Medicine’s 2026 report, which includes contributions from Gabby Kitch of the Yale Center for Natural Carbon Capture, highlights progress and key research needs in advancing ocean alkalinity enhancement as a climate solution.
- Yale physicist Chiara Mingarelli was cited for contributions in pulsar timing array theory and gravitational-wave astrophysics.
- Whether you’re applying for a job, pitching a startup, or hoping to break into the music business, your first evaluator may be an AI system. New research by professor Tristan Botelho and Ph.D. student Qingyang Wang finds that LLMs avoid overt discrimination when evaluating candidates but still rely on racial and gender cues.
- Yale physicist Reina Maruyama was part of the construction and commissioning team for the IceCube experiment, whose principal investigator, Francis Halzen, won the 2026 Nobel Prize for physics.
- The neural networks that power popular AI chatbots process data much differently than humans but capably perform tasks involving language, logic, and math. A new study sheds light on how those systems “think.”
- Mixing minds, stirring conversation.
- Karen Seto’s research has helped reveal how urbanization and other global changes are affecting the planet while pioneering new approaches to geography. The Yale professor recently won the 2026 Vautrin Lud Prize, the field’s highest honor.
Science and Technology, Harvard
- by Sydney BolesWith a split-screen view of a turbulent tech revolution, Michael Brenner makes a case for the value of human guidance
- by Samantha PerfasSherry Turkle’s ‘Artificial Intimacy’ is a book-length response to ‘fundamental lie we’re telling ourselves’
- by Jacob SweetSome say they rob games of color, spontaneity, but others argue changes aren’t always for worse
- by Ryan MulcahyContemplating a future where machines break free of human meddling
- by Sarah LamodiTips for breaking toxic addiction to tech, even when work won’t let you let go
- by gazettebeckycolemanWriter, activist Cory Doctorow wants you to be mad about tech monopoly — but not blindly terrified of AI
- by Sydney BolesHarvard’s history with the prestigious prize may surprise you
- by Elizabeth ZonarichScholars, industry leaders debate how to reduce risks of AI, ensure humans remain in charge of development
- by Sydney BolesIt started with idea to help beleaguered fishermen. Now marine science lab promotes healthy fisheries, trains local workforce, seeks clues to longevity
- by Elizabeth ZonarichDavid Malan says CS50’s AI tutor is a net positive — and getting better
Allen School, Washington
- by Kristine WhiteAllen School Ph.D. student Shangbin Feng aims to build a more open and democratic artificial intelligence future. To that end, his research focuses on model collaboration, where “multiple AI models, trained on different data, by different people, and thus possess diverse skills and strengths, collaborate, compose and complement each other.” In December, Feng was named […]
- by Kristine WhiteAt the recent ACM CHI Conference on Human Factors in Computing Systems (CHI 2026), Allen School researchers brought home multiple accolades for their innovative work in human-computer interaction (HCI) and artificial intelligence. Their projects ranged from interactive systems that allow users to collaborate with AI agents on projects with more flexibility, to an AI-based tool […]
- by Kristine WhiteAs the director of the Allen School’s Makeability Lab, Allen School professor and alum Jon Froehlich (Ph.D., ‘11) utilizes human-computer interaction (HCI) and machine learning to tackle high-impact socially relevant problems. Already, his work has led to improved city planning and sidewalk infrastructure across the globe, and he has developed technologies that enable blind and […]
- by Kristine WhiteAlthough Allen School professor emeritus Richard Ladner started his career as one of the founders of the University of Washington’s theoretical computer science group, he has grown to become a leading researcher in accessible technology and an advocate for expanding access to computer science for students with disabilities.Through mentoring and advocacy, he has directly helped […]
- by Kristine WhiteProgram obfuscation, which aims to obscure the inner workings of a computer program while maintaining its functionality, is a central goal in cryptography and software protection. Allen School Ph.D. student Er-Cheng Tang and collaborator Mi-Ying (Miryam) Huang, a Ph.D. student at the University of Southern California recently developed the first quantum state obfuscation scheme for […]
- by Kristine WhiteAs a computer systems researcher, Allen School professor and alum Ratul Mahajan (Ph.D., ‘05) has helped develop technologies powering the networks that support multiple aspects of modern society — from operating online banking accounts to scrolling social media. The Association for Computing Machinery (ACM) recognized Mahajan among its for his groundbreaking “contributions to network verification […]
- by Kristine WhiteLast month, the Computing Research Association (CRA) recognized a select group of undergraduate students from across North America who have made notable contributions to the field through research. This year’s cohort in the CRA Outstanding Undergraduate Researcher Awards included four Allen School undergraduates — awardee Haoquan Fang, finalist Hao Xu and honorable mention recipients Kaiyuan […]
- by Kristine WhiteAllen School professor Thomas Rothvoss has carved a career out of complexity. As a member of the school’s Theory of Computation Group, Rothvoss examines the theoretical limits of computer algorithms that are designed to analyze large, complex datasets. His aim is to settle long-standing problems in the field of combinatorial optimization — an area where […]
- by Kristine WhiteIn the 1995 movie “Clueless,” lead character Cher Horowitz has a digital closet that allows her to virtually try on outfits. Proving some ideas never go out of style, Allen School professor Ira Kemelmacher-Shlizerman, director of the UW Reality Lab and member of the UW Graphics & Imaging Group (GRAIL), has spent the past two […]
- by Kristine WhiteAs large language models become more popular and widespread, it becomes increasingly important to understand the composition of their training data and how it affects the model’s behavior. The first step is to make these texts searchable. To make searching on such a large scale more efficient and affordable, researchers at the University of Washington […]