Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
SPRINT: Enabling Interleaved Planning and Parallelized Execution in Reasoning ModelsAlphabet · neural information processing systemsJan 1, 2025Reinforcement Learning with Backtracking FeedbackAlphabet, Microsoft · neural information processing systemsJan 1, 2025Marginal-Nonuniform PAC LearnabilityAlphabet · neural information processing systemsJan 1, 2025Quantization-Free Autoregressive Action TransformerGoogle DeepMind · neural information processing systemsJan 1, 2025Closed-Form Training Dynamics Reveal Learned Features and Linear Structure in Word2Vec-like ModelsGoogle DeepMind · neural information processing systemsJan 1, 2025Fast attention mechanisms: a tale of parallelismAlphabet · neural information processing systemsJan 1, 2025Principled Fine-tuning of LLMs from User-Edits: A Medley of Preference, Supervision, and RewardDatabricks, Google DeepMind · neural information processing systemsJan 1, 2025Large language models can learn and generalize steganographic chain-of-thought under process supervisionGoogle DeepMind · neural information processing systemsJan 1, 2025Nearly-Linear Time and Massively Parallel Algorithms for $k$-anonymityAlphabet · neural information processing systemsJan 1, 2025Exact and Linear Convergence for Federated Learning under Arbitrary Client Participation is AttainableAlphabet · neural information processing systemsJan 1, 2025Provable Meta-Learning with Low-Rank AdaptationsAlphabet · neural information processing systemsJan 1, 2025Data-Adaptive Exposure Thresholds under Network InterferenceAlphabet · neural information processing systemsJan 1, 2025VESSA: Video-based objEct-centric Self-Supervised Adaptation for Visual Foundation ModelsAlphabet · neural information processing systemsJan 1, 2025Tiled Flash Linear Attention: More Efficient Linear RNN and xLSTM KernelsGoogle DeepMind · neural information processing systemsJan 1, 2025Universal Sequence PreconditioningAlphabet, Google DeepMind · neural information processing systemsJan 1, 2025SpecEM: Training-Free LLM Ensembling via Iterative Drafting, Verification, and Online FeedbackAlphabet · neural information processing systemsJan 1, 2025HQA-VLAttack: Towards High Quality Adversarial Attack on Vision-Language Pre-Trained ModelsAlphabet · neural information processing systemsJan 1, 2025Sparse Gaussian Processes: Structured Approximations and Power-EP RevisitedGoogle DeepMind · neural information processing systemsJan 1, 2025Weaver: Shrinking the Generation-Verification Gap by Scaling Compute for VerificationAlphabet · neural information processing systemsJan 1, 2025Recursive Inference Scaling: A Winning Path to Scalable Inference in Language and Multimodal SystemsAlphabet, Google DeepMind · neural information processing systemsJan 1, 2025The end of Coding: Predicting a Post-Code World with AI-Native Software DevelopmentAlphabet · International Conference on Interdisciplinary Research in Science, Engineering, and TechnologyJan 1, 2025DoubleDipper: Recycling Contexts for Efficient and Attributed In-Context LearningAlphabet, Google DeepMind · International Joint Conference on Natural Language Processing and Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (IJCNLP-AACL)Jan 1, 2025Item-Language Model: Improving Large Language Model for Recommendation via Item-Language Representation LearningAlphabet · International Joint Conference on Natural Language Processing and Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (IJCNLP-AACL)Jan 1, 2025Image Restoration Through Inversion by Direct Iteration (InDI)Alphabet · Imaging and Applied OpticsJan 1, 2025MLE-Dojo: Interactive Environments for Empowering LLM Agents in Machine Learning EngineeringAlphabet, Google DeepMind · neural information processing systemsJan 1, 2025