Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
Do LLMs Really Need 10+ Thoughts for “Find the Time 1000 Days Later”? Towards Structural Understanding of LLM OverthinkingGoogle DeepMind · Annual Meeting of the Association for Computational Linguistics (ACL)Jan 1SAEdit: Token-Level Control for Continuous Image Editing via Sparse AutoencoderGoogle DeepMind · Lecture notes in computer scienceJan 1Advancing geometry with AI: Multi-agent generation of polytopesGoogle DeepMind · Advances in Theoretical and Mathematical PhysicsJan 1Value Under Ignorance in Universal Artificial IntelligenceGoogle DeepMind · arXiv (Cornell University)Dec 18, 2025Value Under Ignorance in Universal Artificial IntelligenceGoogle DeepMind · arXiv (Cornell University)Dec 18, 2025Quantum oracles give an advantage for identifying classical counterfactualsGoogle DeepMind · arXiv (Cornell University)Dec 15, 2025The price of precision: the cost of preprocessing for automated code revision in code reviewGoogle DeepMind · Empirical Software EngineeringDec 15, 2025Towards a Science of Scaling Agent SystemsAlphabet, Google DeepMind · arXiv (Cornell University)1 citationDec 9, 2025Buffer Centering for bittide Synchronization via Frame RotationGoogle DeepMind · IEEE Conference on Decision and Control (CDC)Dec 9, 2025Towards a Science of Scaling Agent SystemsAlphabet, Google DeepMind · arXiv (Cornell University)Dec 9, 2025Is Generation Required for Data-Efficient Perception?Google DeepMind · arXiv (Cornell University)Dec 9, 2025Navigating with Annealing Guidance Scale in Diffusion SpaceGoogle DeepMind · SIGGRAPH Asia Conference Papers5 citationsDec 8, 2025Improving Streaming ASR via Differentially Private Fusion of Data from Multiple SourcesAlphabet, Google DeepMind, OpenAI · IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)Dec 6, 2025mSTEB: Massively Multilingual Evaluation of LLMs on Speech and Text TasksGoogle DeepMind · IEEE Automatic Speech Recognition and Understanding Workshop (ASRU)Dec 6, 2025WorldPack: Dynamic Frame Compression for Long-context Video World ModelingGoogle DeepMind · arXiv (Cornell University)Dec 2, 2025LLM CHESS: Benchmarking Reasoning and Instruction-Following in LLMs through ChessGoogle DeepMind · arXiv (Cornell University)Dec 1, 2025Selecting User Histories to Generate LLM Users for Cold-Start Item RecommendationAlphabet, Google DeepMind · arXiv (Cornell University)Nov 27, 2025Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving MemoryGoogle DeepMind · arXiv (Cornell University)1 citationNov 25, 2025Budget-Aware Tool Use Enables Effective Agent ScalingGoogle DeepMind · arXiv (Cornell University)1 citationNov 21, 2025Cascading Adversarial Bias from Injection to Distillation in Language ModelsAlphabet, Google DeepMind · ACM SIGSAC Conference on Computer and Communications Security (CCS)2 citationsNov 19, 2025PAS: A Training-Free Stabilizer for Temporal Encoding in Video LLMsGoogle DeepMind · arXiv (Cornell University)1 citationNov 14, 2025Olympiad-level formal mathematical reasoning with reinforcement learningGoogle DeepMind · Nature19 citationsNov 12, 2025SWE-fficiency: Can Language Models Optimize Real-World Repositories on Real Workloads?Alphabet, Google DeepMind · arXiv (Cornell University)1 citationNov 8, 2025Personalized federated learning with exact stochastic gradient descentGoogle DeepMind · Applied Intelligence11 citationsNov 1, 2025Sample-Based Planning and Learning with Function ApproximationGoogle DeepMind · Statistical ScienceNov 1, 2025