Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
EgoReasoner: Learning Egocentric 4D Reasoning via Task-Adaptive Structured ThinkingAlphabet, Google DeepMind · arXiv (Cornell University)Mar 6Free Lunch for Pass@$k$? Low Cost Diverse Sampling for Diffusion Language ModelsGoogle DeepMind · arXiv (Cornell University)Mar 5Dual-Modality Multi-Stage Adversarial Safety Training: Robustifying Multimodal Web Agents Against Cross-Modal AttacksAlphabet, Google DeepMind · arXiv (Cornell University)Mar 4Generalization Properties of Score-matching Diffusion Models for Intrinsically Low-dimensional DataGoogle DeepMind · arXiv (Cornell University)Mar 4Bittide: Control Time, Not FlowsGoogle DeepMind · IEEE MicroMar 4Expected Reward Prediction, with Applications to Model RoutingGoogle DeepMind · arXiv (Cornell University)Mar 3Rich Insights from Cheap Signals: Efficient Evaluations via Tensor FactorizationGoogle DeepMind · arXiv (Cornell University)Mar 2Position: Gaussian DP for Reporting Differential Privacy Guarantees in Machine LearningGoogle DeepMind · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)Mar 1SoK: The Hitchhiker’s Guide to Efficient, End-to-End, and Tight DP AuditingGoogle DeepMind · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)Mar 1Your Privacy Depends on Others: Collusion Vulnerabilities in Individual Differential PrivacyGoogle DeepMind · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)Mar 1Reasoning Introduces New Poisoning Attacks Yet Makes Them More ComplicatedGoogle DeepMind · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)Mar 1Architectural Backdoors for Within-Batch Data Stealing and Model Inference ManipulationGoogle DeepMind · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)Mar 1Differentially Private Adaptation of Diffusion Models via Noisy Aggregated EmbeddingsGoogle DeepMind, Microsoft · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)Mar 1What You Read is What You Classify: Highlighting Attributions to Text and Text-Like InputsGoogle DeepMind · arXiv (Cornell University)Feb 27What You Read is What You Classify: Highlighting Attributions to Text and Text-Like InputsGoogle DeepMind · arXiv (Cornell University)Feb 27Imagining and building wise machines: the centrality of AI metacognitionGoogle DeepMind · Trends in Cognitive Sciences8 citationsFeb 26What Matters for Simulation to Online Reinforcement Learning on Real RobotsGoogle DeepMind · arXiv (Cornell University)Feb 23Deep dive into model-free reinforcement learning for underwater locomotion: theory and practiceGoogle DeepMind · Bioinspiration & BiomimeticsFeb 23OpenSage: Self-programming Agent Generation EngineGoogle DeepMind · arXiv (Cornell University)Feb 18Symmetry in language statistics shapes the geometry of model representationsGoogle DeepMind · arXiv (Cornell University)1 citationFeb 16Simplicity and Complexity in Combinatorial OptimizationGoogle DeepMind · Entropy1 citationFeb 15AI-Assisted Scientific Assessment: A Case Study on Climate ChangeGoogle DeepMind · arXiv (Cornell University)Feb 10Step-resolved data attribution for looped transformersGoogle DeepMind · arXiv (Cornell University)Feb 10Stress-Testing Alignment Audits With Prompt-Level Strategic DeceptionAlphabet, Google DeepMind · arXiv (Cornell University)Feb 9CoRefine: Confidence-Guided Self-Refinement for Adaptive Test-Time ComputeGoogle DeepMind · arXiv (Cornell University)Feb 9