Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
SpectralShift: Effective Context Window Extension of Gated DeltaNet via Spectral ReparameterizationMicrosoft · arXiv (Cornell University)Sep 13OpWeave: Flexible Operator Disaggregation for Heterogeneous LLM ServingAlphabet · arXiv (Cornell University)Sep 13Private Graph Property TestingAlphabet · arXiv (Cornell University)Sep 13MoUNets: Static-Weighted Ensemble of UNet-Based Experts for Levator Hiatus Segmentation in Pelvic Floor UltrasoundAlibaba · Ultrasound in Medicine & BiologySep 13ActiveRDMA: SmartNIC-offloaded, Multi-Step RDMA operationsNVIDIA · Proceedings of the International Conference on Parallel ProcessingSep 13RapidGEMM: An Efficient Tile-agnostic GEMM for Mixture-of-Experts TrainingMicrosoft · Proceedings of the International Conference on Parallel ProcessingSep 13Evaluating Large Language Models for Antisemitic Incident ClassificationMicrosoft · Digital Hate ReviewSep 13JaxAHT: A JAX-Based Library for Ad Hoc TeamworkAlphabet · arXiv (Cornell University)Sep 12IBBench-Light: A Paired Evaluation of Task-Conditioned Responses to External DirectivesAlphabet · arXiv (Cornell University)Sep 12JumpStart Your Policy Learning with Lessons from 160,000 Training RunsMicrosoft · arXiv (Cornell University)Sep 12HarnessBandit: Joint Learnability-Transferability Scheduling for Multi-Harness Agentic Reinforcement LearningAlibaba · arXiv (Cornell University)Sep 12Does Reasoning Improve Psychological Depth in Large Language Models? It Depends on Who's JudgingAlphabet · arXiv (Cornell University)Sep 12Certifying Model Upgrades with Slice-Wise Non-Regression and Incumbent FallbackAlibaba · arXiv (Cornell University)Sep 12From Token Probabilities to Semantic Constraints: Towards Declarative Probabilistic Evaluation of Language ModelsAllen Institute for AI · arXiv (Cornell University)Sep 11Learning to Solve Hard Problems in RL for LLMs by Never Giving UpAllen Institute for AI · arXiv (Cornell University)Sep 11AutoTailor: Automatic, User-Aligned Capability Selection and Adaptation for Web AgentsMicrosoft · arXiv (Cornell University)Sep 11How User-AI Mistreatment Occurs and Matters in Conversational Systems?Microsoft · arXiv (Cornell University)Sep 11BOOST: Concurrent Access to Host Memory and HBM to Accelerate LLM InferenceNVIDIA · arXiv (Cornell University)Sep 11RoofLang: Enabling AI-Driven Architecting of LLM Inference SystemsMicrosoft · arXiv (Cornell University)Sep 11SCOPE-OPSD: Fisher-Conditioned Privileged Subspaces for On-Policy Self-DistillationAlibaba · arXiv (Cornell University)Sep 11Rethinking Heterogeneous System Disaggregation for Subquadratic AttentionNVIDIA · arXiv (Cornell University)Sep 11Differential Privacy Meets Fixed Parameter Tractability: Algorithms and Lower BoundsAlphabet · arXiv (Cornell University)Sep 11Composing Detector Error ModelsNVIDIA · arXiv (Cornell University)Sep 11Temporal Recurrence Favors Fewer LayersSakana AI · arXiv (Cornell University)Sep 11MIMA: Multi-Interest Recommendation via Multi-Positive Exclusive AssignmentAlibaba · arXiv (Cornell University)Sep 11