The day's stories
BenchmarkITBench-AA: Frontier Models Score Below 50% on the First Benchmark for Agentic Enterprise IT Tasks — by Artificial Analysis and IBMArtificial Analysis and IBM have introduced ITBench-AA, a pioneering benchmark designed to assess frontier AI models in the context of agentic enterprise IT tasHugging Face BlogMay 27· 60%FundingIn more good news for Amazon, Snowflake signs $6B deal with AWS for AI CPU chipsSnowflake has entered into a significant $6 billion agreement with Amazon Web Services (AWS) to utilize AWS's ARM-based Graviton CPU chips for enhancing its AI TechCrunch AIMay 27· 59%Product LaunchElevenLabs' new music-generation model can switch genres mid-trackElevenLabs has unveiled Music v2, a cutting-edge music-generation model that allows artists to seamlessly switch genres within a single track. This innovative fTechCrunch AIMay 27· 58%RegulationChina is increasingly keeping its best AI talent to itselfChina has begun to impose stricter travel restrictions on its leading AI researchers and executives, mandating government approval for any international travel.TechCrunch AIMay 27· 58%ResearchShipping a Trillion Parameters With a Hub Bucket: Delta Weight Sync in TRLRecent advancements in reinforcement learning training have been made with the introduction of a delta weight synchronization method. This innovative approach aHugging Face BlogMay 27· 58%Product LaunchSOND, a sleep tech startup from Bose's former head of sleep, exits stealth with $7MSOND, a new sleep tech startup founded by Yadid Ayzenberg, the former head of sleep at Bose, has officially exited stealth mode after securing $7 million in funTechCrunch AIMay 27· 58%Product LaunchPayroll startup Remote says it grew revenue 50% per employee without adding headcountRemote, a payroll startup, has reported a remarkable 50% increase in revenue per employee, achieving this growth without adding to its workforce. This significaTechCrunch AIMay 27· 57%Product LaunchRobinhood now lets your AI agents trade stocksRobinhood has unveiled a new feature that enables users to delegate stock trading to AI agents. This innovative approach allows users to set up dedicated accounTechCrunch AIMay 27· 57%RegulationWebsites have a new way to spy on visitors: analyzing their SSD activityA new tracking technique known as FROST has emerged, allowing websites to monitor visitors' SSD activity. This method enables sites to spy on other websites beiArs Technica AIMay 27· 57%OtherTech CEOs are apparently suffering from AI psychosisRecent reports indicate that tech CEOs are experiencing a phenomenon dubbed 'AI psychosis,' where they overestimate the capabilities of artificial intelligence.TechCrunch AIMay 27· 56%Product LaunchMeta launches Instagram, Facebook, and WhatsApp subscriptions, with more to come, including AI plansMeta has introduced subscription plans for its popular platforms, including Instagram, Facebook, and WhatsApp. These subscriptions aim to provide users with enhTechCrunch AIMay 27· 56%ResearchYour SEO strategy is optimized for a search engine that no longer exists.Recent changes in Google's search algorithms have fundamentally altered the landscape of search engine optimization (SEO). With the rise of AI-generated answersTechCrunch AIMay 27· 56%FundingAI coding startup Cognition raises $1B at $25B pre-money valuationCognition, an AI coding startup, has successfully raised over $1 billion in its latest funding round, achieving a staggering pre-money valuation of $25 billion.TechCrunch AIMay 27· 56%RegulationYouTube will now automatically label AI videosYouTube has announced a new feature that will automatically label videos utilizing significant photorealistic AI technology. This initiative is part of the platTechCrunch AIMay 27· 55%ResearchThe Constraint Tax: Measuring Validity-Correctness Tradeoffs in Structured Outputs for Small Language ModelsA recent paper published on arXiv introduces the concept of 'constraint tax', which quantifies the trade-offs between validity and correctness in structured outarXiv cs.LGMay 27· 64%ResearchGEM: Geometric Entropy Mixing for Optimal LLM Data CurationA new research framework called Geometric Entropy Mixing (GEM) has been introduced to optimize data curation for large language models (LLMs). This innovative aarXiv cs.LGMay 27· 64%ResearchAnchor: Mitigating Artifact Drift in Agent Benchmark GenerationResearchers have introduced a new task-generation pipeline called Anchor, aimed at addressing the issue of artifact drift in the training and evaluation of AI aarXiv cs.AIMay 27· 64%BenchmarkJobBench: Aligning Agent Work With Human WillResearchers have unveiled JobBench, a novel benchmark designed to evaluate AI agents based on their effectiveness in assisting with high-priority tasks across aarXiv cs.AIMay 27· 64%ResearchPersonalizing Embodied Multimodal Large Language Model Agents over Long-term User InteractionsRecent research introduced POLAR, a novel framework designed for personalizing embodied multimodal large language model agents through long-term user interactioarXiv cs.AIMay 27· 63%ResearchARBITER: Reasoning Trajectory Basins and Majority Vote Failures in Test-Time SamplingResearchers have introduced ARBITER, a novel model-agnostic method designed to address the issue of wrong-majority failures in test-time sampling of language moarXiv cs.LGMay 27· 63%BenchmarkOmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief ModelingThe introduction of OmniToM marks a significant advancement in the evaluation of Theory of Mind (ToM) capabilities in large language models (LLMs). This new benarXiv cs.AIMay 27· 63%ResearchNeural Bayesian Sequential RoutingResearchers have unveiled a novel framework known as Neural Bayesian Sequential Routing (NBSR), which enhances neural inference by conceptualizing it as active arXiv cs.LGMay 27· 63%ResearchPlanning Neural Dynamics with Lie Group Embedding through Supervised Projective Manifold LearningA recent study has introduced a novel approach called LieEDNN, which utilizes Lie group embeddings to enhance the stability and learnability of dynamical neuralarXiv cs.LGMay 27· 63%BenchmarkConstraint acquisition needs better benchmarksRecent advancements in the field of Constraint Acquisition (CA) have highlighted the need for improved benchmarking methods. Researchers have introduced MPMMinearXiv cs.AIMay 27· 63%ResearchExperiments in Agentic AI for ScienceRecent advancements in agentic AI have been explored through two innovative frameworks aimed at enhancing scientific workflows. The first framework, DeepTS/DeeparXiv cs.AIMay 27· 62%BenchmarkYour Agents Are Aging Too: Agent Lifespan Engineering for Deployed SystemsA recent study has introduced AgingBench, a benchmark designed to evaluate the reliability and lifespan of AI agents after they have been deployed. This researcarXiv cs.AIMay 27· 62%ResearchAirCast-SR: A Foundation Model for Kilometer-Scale Atmospheric Super-Resolution via Latent Consistency DiffusionResearchers have introduced AirCast-SR, a groundbreaking foundation model designed to enhance global weather forecasting capabilities. By employing a three-dimearXiv cs.LGMay 27· 62%ResearchIs Agent Memory a Database? Rethinking Data Foundations for Long-Term AI Agent MemoryRecent research has introduced a novel framework called Governed Evolving Memory (GEM) aimed at enhancing the memory capabilities of long-term AI agents. This farXiv cs.AIMay 27· 62%ResearchOn the Push-Based Asynchronous Federated Learning: A Bias-Correction Aggregation ApproachRecent research has unveiled a novel framework called PushCen-ADFL, designed to enhance asynchronous decentralized federated learning. This innovative approach arXiv cs.LGMay 27· 62%ResearchTSFMAudit: Data Contamination Auditing in Forecasting Time Series Foundation ModelsResearchers have developed a novel method called TSFMAudit, aimed at auditing data contamination in forecasting time series foundation models (TSFMs). This new arXiv cs.LGMay 27· 61%ResearchSilIF: Silhouette-Augmented Isolation Forest for Unsupervised Transaction Fraud DetectionThe recent paper on SilIF presents a novel approach to unsupervised transaction fraud detection by enhancing the traditional Isolation Forest method. This new marXiv cs.LGMay 27· 61%ResearchWhen Rule Violations Are Rare: Chimera Training for Logical Anomaly DetectionA recent study published on arXiv has introduced a groundbreaking approach called chimera training, specifically designed for logical anomaly detection in envirarXiv cs.LGMay 27· 60%ResearchCan LLMs Introspect? A Reality CheckRecent research has critically examined the introspective capabilities of large language models (LLMs). The study highlights that LLMs tend to misinterpret chanarXiv cs.AIMay 27· 60%