The day's stories
Product LaunchCheaper, faster, and culturally aware, Avataar's video AI is built for India's scaleAvataar AI has introduced a new video AI model named Varya, specifically tailored for the Indian market. This model is designed to be culturally aware, understaTechCrunch AIJun 12· 59%Open Sourceolmo-eval: An evaluation workbench for the model development loopHugging Face has unveiled olmo-eval, a new evaluation workbench aimed at improving the model development loop specifically for large language models (LLMs). ThiHugging Face BlogJun 12· 59%OtherMeta's months-old AI unit is a soul-crushing gulag, say the engineers stuck inside itMeta's Applied AI team is currently experiencing significant internal strife, as employees voice their dissatisfaction with the working conditions and managemenTechCrunch AIJun 12· 58%FundingJeff Bezos's Prometheus raises $12B to build an 'artificial general engineer' for the physical worldPrometheus, a startup co-founded by Jeff Bezos, has successfully raised $12 billion in funding to develop an 'artificial general engineer' designed to automate TechCrunch AIJun 12· 57%RegulationGoogle sues alleged Chinese cybercrime operation that used AI to send scam textsGoogle has initiated legal action against Outsider Enterprise, a Chinese cybercrime group accused of leveraging artificial intelligence to distribute fraudulentTechCrunch AIJun 12· 57%FundingIt’s hot IPO summer, and the MANGOS are ripeThe IPO market is witnessing a notable resurgence, characterized by the emergence of a new group of tech companies collectively referred to as MANGOS, which incTechCrunch AIJun 12· 56%FundingMistral is rumored to be raising €3B at €20B valuationMistral AI is reportedly in advanced discussions to secure €3 billion in funding, which would elevate its valuation to around €20 billion. This potential fundinTechCrunch AIJun 12· 56%SafetyPeopleSoft 0-day affecting hundreds of organizations steals gigabytes of dataA critical zero-day vulnerability in Oracle's PeopleSoft software, identified as CVE-2026-35273, has been exploited by the ransomware group ShinyHunters, impactArs Technica AIJun 12· 55%ResearchPythagoras-Prover: Advancing Efficient Formal Proving via Augmented Lean FormalisationThe Pythagoras-Prover represents a significant advancement in the field of formal proving, leveraging augmented Lean formalisation techniques to enhance efficiearXiv cs.AIJun 12· 65%ResearchArbor: Tree Search as a Cognition Layer for Autonomous AgentsThe recent introduction of Arbor, a multi-agent framework, marks a significant advancement in the field of autonomous agents. By implementing a structured tree arXiv cs.AIJun 12· 64%ResearchTrajGenAgent: A Hierarchical LLM Agent for Human Mobility Trajectory GenerationResearchers have introduced TrajGenAgent, a novel hierarchical large language model (LLM) designed to generate realistic human mobility trajectories. This innovarXiv cs.AIJun 12· 64%ResearchPermDoRA -- Understanding Adapter Interference in Language Models: Limits of Parameter-Space GeometryThe recent study titled 'PermDoRA' delves into the phenomenon of adapter interference in large language models (LLMs). It tests the hypothesis that enforcing orarXiv cs.LGJun 12· 64%ResearchEvoflux: Inference-Time Evolution of Executable Tool Workflows for Compact AgentsThe introduction of Evoflux marks a significant advancement in the field of AI, particularly for compact language models. This new evolutionary search method enarXiv cs.AIJun 12· 64%Open SourceToolSense: A Diagnostic Framework for Auditing Parametric Tool Knowledge in LLMsToolSense has been introduced as an open-source diagnostic framework aimed at auditing the parametric tool knowledge of large language models (LLMs). This framearXiv cs.AIJun 12· 63%ResearchLoss Landscape Diagnosis for Gradient-Based Gray-Scott System Inversion: Disentangling the Roles of PINN ComponentsA recent study delves into the challenges faced by physics-informed neural networks (PINNs) when applied to the Gray-Scott reaction-diffusion system for parametarXiv cs.LGJun 12· 63%ResearchPersonaDrive: Human-Style Retrieval-Augmented VLA Agents for Closed-Loop Driving SimulationThe introduction of PersonaDrive marks a significant advancement in the field of autonomous driving simulations. This innovative method leverages a vision-languarXiv cs.AIJun 12· 63%ResearchTo Intervene or Not: Guiding Inference-time Alignment with Probabilistic Model BlendingRecent research has introduced BlendIn, a novel inference-time alignment framework designed to enhance the performance of large language models. This framework arXiv cs.LGJun 12· 62%Research"Did you lie?" Evaluating Lie Detectors across Model Scale and Belief-Verified Model OrganismsA recent study published on arXiv has introduced a novel approach to evaluating lie detectors in AI by utilizing 13 reasoning model organisms and a prompted-lyiarXiv cs.AIJun 12· 62%ResearchFew-Shot Resampling for Scalable Statistically-Sound Data MiningThe recent paper on FewRS presents a groundbreaking method for enhancing the efficiency of data mining through a novel resampling technique. This approach allowarXiv cs.LGJun 12· 62%ResearchMechanical Field Networks: Structured Neural Dynamics for Multivariate SystemsResearchers have introduced a novel model called Mechanical Field Networks (MF-Net), which is designed to effectively represent multivariate systems through a rarXiv cs.LGJun 12· 62%ResearchStrategic Decision Support for AI AgentsA recent study has introduced a novel framework aimed at enhancing strategic decision support for AI agents. This framework focuses on optimizing the reliabilitarXiv cs.AIJun 12· 62%ResearchProHiFlo: Hierarchical Flow Matching with Functional Guidance for De Novo Protein GenerationRecent advancements in protein generation have been marked by the introduction of ProHiFlo, a novel hierarchical flow matching framework. This innovative approaarXiv cs.LGJun 12· 61%ResearchRestless bandits with imperfect binary feedback: PCL-indexability analysis and computationA recent paper published on arXiv explores the complexities of restless bandits that operate under binary latent states and receive imperfect feedback. The autharXiv cs.LGJun 12· 61%ResearchDeployment-Centered Evaluation: Predicting Query-Level Rejection Risk in a Clinical LLM SystemA recent study has introduced a classifier designed to predict user rejection of responses generated by large language models (LLMs) in clinical environments. TarXiv cs.AIJun 12· 61%ResearchPhysics-informed generative AI for semiconductor manufacturing: Enforcing hard physical constraints in generative models by constructionRecent research highlights the importance of integrating hard physical constraints into generative AI models specifically for semiconductor manufacturing. InstearXiv cs.LGJun 12· 60%ResearchDual-Stance Evaluation of Sycophancy: The Structure of Agreement and the Limits of InterventionA recent study published on arXiv presents a novel approach called dual-stance evaluation to analyze the behavior of large language models (LLMs) in relation toarXiv cs.LGJun 12· 60%ResearchFrom AGI to ASIThe recent report titled 'From AGI to ASI' explores the critical transition from artificial general intelligence (AGI) to artificial superintelligence (ASI). ItarXiv cs.AIJun 12· 59%ResearchBernstein-Schur Kernels: Random Features by Sketched Modulation and Radial RandomizationA new research paper has introduced Bernstein-Schur kernels, which represent a significant advancement in the field of kernel methods for machine learning. ThesarXiv cs.LGJun 12· 58%