The day's stories
RegulationA tech worker-backed PAC is bringing a $5M knife to Big Tech’s $100M gunfightThe Guardrails Alliance, a newly established super PAC backed by tech workers, is entering the political arena with a focus on promoting responsible AI legislatTechCrunch AIJun 18· 58%OtherAlmost half of U.S. singles feel negatively about AI in dating, Match saysA recent study conducted by Match Group highlights a significant skepticism among U.S. singles regarding the role of AI in dating. The survey found that 47% of TechCrunch AIJun 18· 58%OtherOpenAI is bringing on some big guns in the lead-up to its IPOIn a strategic move ahead of its anticipated IPO, OpenAI has expanded its leadership team by hiring Noam Shazeer, a key contributor to generative AI and the TraTechCrunch AIJun 18· 57%Product LaunchSnap spins off AI video team into new company, Dotmo, due to costsSnap has officially spun off its generative AI video team into a new entity named Dotmo. This strategic move aims to address the rising costs associated with maTechCrunch AIJun 18· 57%RegulationAI data centers just got a government-mandated fast lane to the gridThe Federal Energy Regulatory Commission (FERC) has introduced a new regulation that requires grid operators to expedite the interconnection process for AI dataTechCrunch AIJun 18· 57%ResearchMosaicLeaks: Can your research agent keep a secret?The recent article from Hugging Face discusses the privacy risks that arise when deep research agents integrate private documents with external web searches. ThHugging Face BlogJun 18· 57%Product LaunchAmazon hopes to challenge Nvidia more directly by selling its AI chipsAmazon Web Services (AWS) is taking a significant step in the AI chip market by considering the sale of its proprietary AI chip, Trainium, to other companies. TTechCrunch AIJun 18· 57%FundingGeneral Intuition in talks to raise $300M at around $2B valuationGeneral Intuition, a startup specializing in AI agents for spatial-temporal reasoning, is currently in negotiations to secure $300 million in funding. This new TechCrunch AIJun 18· 56%FundingAI inference startup Baseten reportedly raising $1.5B months after its last mega-roundAI inference startup Baseten is on the verge of securing an impressive $1.5 billion in funding, which would elevate its valuation to $13 billion. This comes jusTechCrunch AIJun 18· 55%Product Launch'Queer Eye's' life coach Karamo Brown launches Kē, a wellness app featuring his AI digital cloneKaramo Brown, known for his role as a life coach on 'Queer Eye', has introduced Kē, a new wellness app designed to enhance personal development. The app integraTechCrunch AIJun 18· 54%RegulationHow to turn off AI in your Google DocsGoogle has released a guide for users on how to disable the AI feature known as 'Gemini' in Google Docs. This feature, which integrates AI capabilities into theTechCrunch AIJun 18· 54%ResearchBreaking the Solver Bottleneck: Training Task Generators at the Learnable FrontierA recent study has introduced PROPEL, a novel framework designed to enhance task generation for reinforcement learning. This advancement aims to alleviate the sarXiv cs.LGJun 18· 65%ResearchSEAGym: An Evaluation Environment for Self-Evolving LLM AgentsThe recent introduction of SEAGym marks a significant advancement in the evaluation of self-evolving large language model (LLM) agents. This new environment aimarXiv cs.AIJun 18· 64%BenchmarkCan LLMs Be CEOs? Benchmarking Strategic Resource Reallocation with Multi-Role Agent SimulationResearchers have developed a new benchmark called CEO-Bench, aimed at evaluating the decision-making capabilities of large language models (LLMs) in the contextarXiv cs.AIJun 18· 64%ResearchAttribution-Guided and Coverage-Maximized Pruning for Structural MoE CompressionRecent research has introduced a novel structural pruning framework specifically designed for Mixture-of-Experts (MoE) models. This framework focuses on maximizarXiv cs.LGJun 18· 64%ResearchDRIFT: Refining Instruction Data via On-Policy Data AttributionThe recent paper on DRIFT presents a novel approach to refining instruction data aimed at improving the performance of Large Language Models (LLMs) during SuperarXiv cs.LGJun 18· 63%ResearchCODEBLOCK: Learning to Supervise Code at the Right GranularityResearchers have introduced a novel framework called CodeBlock, aimed at improving the supervised fine-tuning of code language models. This framework enhances tarXiv cs.LGJun 18· 63%ResearchIncumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation SystemsA recent study published on arXiv investigates the impact of brand dynamics on product recommendations generated by large language models (LLMs), specifically farXiv cs.AIJun 18· 63%ResearchClosing the Feedback Loop: From Experience Extraction to Insight Governance in Verbal Reinforcement LearningRecent research has introduced a groundbreaking three-layer architecture for verbal reinforcement learning, which allows agents to learn from feedback while maiarXiv cs.AIJun 18· 63%ResearchSurrogate Assisted Pedestrian Protection Design via a Foundation Model Orchestrated WorkflowRecent research has introduced an innovative AI-driven workflow aimed at improving pedestrian protection in vehicle crash safety design. By utilizing a foundatiarXiv cs.AIJun 18· 62%ResearchGaussian Mixture Attention: Linear-Time Sequence Mixing via Probabilistic Latent RoutingRecent research has introduced a novel attention mechanism known as Gaussian Mixture Attention (GMA), which significantly improves the scalability of TransformearXiv cs.LGJun 18· 62%ResearchTRIDENT: Breaking the Hybrid-Safety-Physics Coupling for Provably Safe Multi-Agent Reinforcement LearningResearchers have developed TRIDENT, an innovative multi-agent reinforcement learning framework designed to tackle the complexities of hybrid discrete-continuousarXiv cs.LGJun 18· 62%ResearchSAGE: Retain-Aware Post-Hoc Sanitization of Final Unlearning VectorResearchers have introduced a new method called SAGE, which focuses on the post-hoc sanitization of final unlearning vectors in large language models. This innoarXiv cs.LGJun 18· 62%ResearchArtemis: Anatomy-Resolved inTervention for Eliminating Multimodal NeuroImage confounderSResearchers have introduced Artemis, a novel framework designed to enhance graph neural networks (GNNs) by specifically targeting demographic confounders in mularXiv cs.LGJun 18· 62%ResearchFisher Width: A Geometric Measure of Complexity on Statistical ManifoldsA recent paper published on arXiv presents the concept of Fisher width, a novel geometric measure designed to assess complexity within statistical manifolds. TharXiv cs.LGJun 18· 61%ResearchLLM-as-Judge in Education: A Curriculum-Grounded Marking PipelineA recent paper has introduced a novel LLM-as-Judge pipeline aimed at automating the marking of educational assessments, particularly in high-stakes environmentsarXiv cs.AIJun 18· 61%ResearchDissecting model behavior through agent trajectoriesA recent study introduced the Simple Strands Agent (SSA), a novel approach designed to analyze and improve the behavior of AI models by focusing on agent trajecarXiv cs.AIJun 18· 61%BenchmarkMapSatisfyBench: Benchmarking Satisfaction-Aware Map Agents through Behavior-Grounded Implicit Decision FactorsResearchers have developed a new benchmarking framework called MapSatisfyBench, which is designed to evaluate satisfaction-aware map agents. This benchmark focuarXiv cs.AIJun 18· 61%InfrastructureDeepInsight: A Unified Evaluation Infrastructure Across the Physical AI StackDeepInsight has been introduced as a comprehensive evaluation infrastructure aimed at enhancing the assessment of the Physical AI stack. This innovative framewoarXiv cs.AIJun 18· 61%ResearchA Link between Shock-wave Theory and Symmetry-reduced Stochastic Gradient Descent for Artificial Neural NetworksRecent research has unveiled a novel mathematical connection between shock-wave theory and the learning dynamics of stochastic gradient descent (SGD) in artificarXiv cs.LGJun 18· 60%