Research
A paper is listed when one of its authors is affiliated with the lab. Data: OpenAlex.
CodeClinic: Evaluating Automation of Coding Skills for Clinical Reasoning AgentsMicrosoft · arXiv (Cornell University)May 10How Much is Brain Data Worth for Machine Learning?Allen Institute for AI · arXiv (Cornell University)May 10Online Steiner Forest with RecourseMicrosoft · arXiv (Cornell University)May 10PDEAgent-Bench: A Multi-Metric, Multi-Library Benchmark for PDE Solver GenerationAlibaba · arXiv (Cornell University)May 10SkillGen: Verified Inference-Time Agent Skill SynthesisMicrosoft · arXiv (Cornell University)May 9CoDistill-GRPO: A Co-Distillation Recipe for Efficient Group Relative Policy OptimizationAlphabet · arXiv (Cornell University)May 9Generating Leakage-Free Benchmarks for Robust RAG EvaluationMicrosoft · arXiv (Cornell University)May 9FraudBench: A Multimodal Benchmark for Detecting AI-Generated Fraudulent Refund EvidenceAlibaba · arXiv (Cornell University)May 9DARE: Difficulty-Adaptive Reinforcement Learning with Co-Evolved Difficulty EstimationAlphabet · arXiv (Cornell University)May 9The Cancellation Hypothesis in Critic-Free RL: From Outcome Rewards to Token CreditsAlibaba · arXiv (Cornell University)May 9UserGPT Technical ReportAlibaba · arXiv (Cornell University)May 9Token Economics for LLM Agents: A Dual-View Study from Computing and EconomicsAlibaba · arXiv (Cornell University)May 9Effective registration-free dual-phase segmentation for pancreas and pancreatic mass via symmetrical selective feature integrationAlibaba · Medical Image AnalysisMay 9GPUArmor: A Hardware-Software Co-Design for Efficient and Scalable Memory Safety on GPUsNVIDIA · ACM Transactions on Architecture and Code OptimizationMay 9A Hierarchical End-of-Turn Model with Primary Speaker Segmentation for Real-Time Conversational AIMeta · IEEE Conference on Artificial Intelligence (CAI)1 citationMay 8Post-training makes large language models less human-likeAlphabet · arXiv (Cornell University)1 citationMay 8A System-Level Taxonomy of Failure Modes in Large Language Model ApplicationsMicrosoft · IEEE Conference on Artificial Intelligence (CAI)May 8Table-First Guarded Reasoning: Outperforming Larger MLLMs in Chart Question AnsweringMicrosoft · IEEE Conference on Artificial Intelligence (CAI)May 8The Evolution of Kubernetes Architecture for General-Purpose AI Systems (GPAIS)Alphabet · IEEE Conference on Artificial Intelligence (CAI)May 8PREFENSE: A Capability-Aware Benchmark for Evaluating Contextual-Grounded Prompting in LLM-Based Security InvestigationMicrosoft · IEEE Conference on Artificial Intelligence (CAI)May 8PaT: Planning-after-Trial for Efficient Test-Time Code GenerationMicrosoft · arXiv (Cornell University)May 8The Geometric Structure of Models Learning Sparse DataAlphabet · arXiv (Cornell University)May 8Mid-Training with Self-Generated Data Improves Reinforcement Learning in Language ModelsAlphabet, Google DeepMind · arXiv (Cornell University)May 8Human-Inspired Memory Architecture for LLM AgentsMicrosoft · arXiv (Cornell University)May 8Can Revealed Preferences Clarify LLM Alignment and Steering?Microsoft · arXiv (Cornell University)May 8