The day's stories
RegulationOpenAI limits GPT-5.6 rollout after government request, says restrictions shouldn’t be the normOpenAI has decided to limit the rollout of its GPT-5.6 models to a select group of trusted partners, following a request from the U.S. government. This move hasTechCrunch AIJun 26· 59%Product LaunchOpenAI's Jalapeño chip is Big Tech's spiciest move away from NvidiaOpenAI has unveiled its new custom inference chip, Jalapeño, developed in partnership with Broadcom. This launch signifies a pivotal moment in the tech industryTechCrunch AIJun 26· 57%RegulationIt's not about Anthropic vs. OpenAI anymoreRecent developments in AI regulation have shifted the landscape, particularly affecting major players like Anthropic and OpenAI. The U.S. government's increasinTechCrunch AIJun 26· 57%Product LaunchOpenAI poaches Uber India chief to lead its biggest market outside the USOpenAI has made a significant move to strengthen its presence in India by appointing Prabhjeet Singh, the former president of Uber India, as its first managing TechCrunch AIJun 26· 56%ResearchAlgoEvolve: LLM-driven Meta-evolution of Algorithmic Trading ProgramsResearchers have introduced AlgoEvolve, a novel framework that leverages large language models (LLMs) to autonomously evolve algorithmic trading programs. This arXiv cs.AIJun 26· 63%ResearchKG-TRACE: A Neuro-Symbolic Framework for Mechanistic Grounding in Antimicrobial Resistance PredictionThe recent introduction of KG-TRACE marks a significant advancement in the field of antimicrobial resistance (AMR) prediction. This neuro-symbolic framework comarXiv cs.LGJun 26· 62%ResearchPhysics-guided Convolutional Neural Network for Domain Growth Prediction in Systems with Conserved KineticsA team of researchers has introduced a novel physics-guided convolutional neural network (CNN) aimed at predicting the microstructural evolution of binary mixtuarXiv cs.LGJun 26· 62%ResearchDetecting and Controlling Sycophancy with Cascading Linear FeaturesA recent paper published on arXiv introduces a novel method for detecting and controlling sycophancy in language models. The authors propose an iterative data garXiv cs.AIJun 26· 62%BenchmarkLife After Benchmark Saturation: A Case Study of CORE-BenchThe introduction of CORE-Bench v1.1 and CORE-Bench OOD marks a significant advancement in evaluating AI agent performance beyond traditional accuracy metrics. RarXiv cs.AIJun 26· 62%ResearchAccelerating Skill Assessment in Chess: A Drift-Diffusion-Enhanced Elo Rating SystemA new chess skill assessment framework called the Drift-Diffusion-Enhanced Elo Rating System (DD-Elo) has been developed by researchers. This innovative model larXiv cs.AIJun 26· 62%ResearchNecessary but Not Sufficient: Temperature Control and Reproducibility in LLM-as-Judge Safety EvaluationsA recent study published on arXiv examines the reliability of large language models (LLMs) when used as judges in safety evaluations. The research specifically arXiv cs.LGJun 26· 62%ResearchImplementation of reinforcement learning in chemical reaction networks: application to phototaxis as curiosity-driven explorationRecent research has introduced a novel framework that integrates reinforcement learning with chemical reaction networks, specifically applied to the phenomenon arXiv cs.LGJun 26· 62%ResearchKnowledge-augmented Agentic AI for Mental Health Medication Information SeekingA new framework has been developed by researchers to enhance the understanding and safety of antidepressant use through a knowledge-augmented agentic AI system.arXiv cs.AIJun 26· 61%RegulationGoverning Actions, Not Agents: Institutional Attestation as a Governance Model for Autonomous AI SystemsA recent paper proposes a novel governance model for autonomous AI systems that emphasizes the importance of regulating actions rather than the agents behind tharXiv cs.AIJun 26· 61%ResearchAgentic Analysis for Agentic Infrastructure: An LLM-Powered Pipeline for Comparative Governance of DAO and Corporate AI ProtocolsRecent research has introduced an innovative LLM-powered pipeline aimed at analyzing governance structures for AI agent protocols, specifically comparing the ERarXiv cs.AIJun 26· 61%ResearchThe Verification Horizon: No Silver Bullet for Coding Agent RewardsRecent research has shed light on the increasing difficulties in verifying the outputs of advanced coding agents. As these agents improve in their ability to gearXiv cs.AIJun 26· 61%ResearchCOrigami: An AI Pipeline for Co-Designing Flat-Foldable Visually Recognisable OrigamiThe recent introduction of COrigami marks a significant advancement in the intersection of artificial intelligence and art. This innovative AI pipeline is speciarXiv cs.AIJun 26· 61%Research\chisao{}: A GPU-Native Parallel Optimizer for Multimodal Black-Box Functions via Convergence-Anticonvergence OscillationResearchers have unveiled a new optimization algorithm called extit{chisao}, designed specifically for multimodal black-box functions. This GPU-native optimizearXiv cs.LGJun 26· 60%ResearchStatistical and Structural Approaches to Algorithmic FairnessRecent research has focused on the need for algorithmic fairness in machine learning systems, which have evolved into complex socio-technical architectures. ThearXiv cs.LGJun 26· 60%ResearchTopology-Informed Neural Networks for Flood Detection in Optical and Synthetic Aperture Radar ImageryRecent research has introduced topology-informed neural networks designed to significantly improve flood detection capabilities in both optical and synthetic aparXiv cs.LGJun 26· 60%ResearchRefusal Lives Downstream of Persona in Chat ModelsA recent study published on arXiv investigates the relationship between refusal responses and persona traits in instruction-tuned chat models. The research specarXiv cs.AIJun 26· 58%