The day's stories
Model ReleaseSandboxAQ brings its drug discovery models to Claude — no PhD in computing requiredSandboxAQ has announced a significant partnership with Anthropic to integrate its advanced drug discovery AI models into Claude, Anthropic's conversational AI pTechCrunch AIMay 18· 60%InfrastructureFine-Tuning NVIDIA Cosmos Predict 2.5 with LoRA/DoRA for Robot Video GenerationNVIDIA has successfully fine-tuned its Cosmos Predict 2.5 model by employing LoRA (Low-Rank Adaptation) and DoRA (Dynamic Low-Rank Adaptation) techniques, signiHugging Face BlogMay 18· 60%FundingThe Open Agent LeaderboardThe Open Agent Leaderboard has been introduced as a new framework for evaluating general-purpose AI agents. This initiative focuses on measuring the performanceHugging Face BlogMay 18· 59%OtherPaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers BackendPaddleOCR has announced the release of version 3.5, which introduces the ability to run Optical Character Recognition (OCR) and document parsing tasks using theHugging Face BlogMay 18· 58%Product LaunchAmazon's new Alexa+ powered feature can generate podcast episodesAmazon has introduced a groundbreaking feature for its Alexa+ platform, enabling users to generate podcast episodes on demand. Dubbed 'Alexa Podcasts,' this innTechCrunch AIMay 18· 57%OtherSouth Korea's LetinAR is building optics behind AI glassesLetinAR, a South Korean startup, is making significant strides in the development of optical technology for AI smart glasses. With a focus on creating lighter, TechCrunch AIMay 18· 57%FundingAnthropic has acquired the dev tools startup used by OpenAI, Google, and CloudflareAnthropic has made a strategic move by acquiring Stainless, a startup that specializes in automating the creation of software development kits (SDKs) utilized bTechCrunch AIMay 18· 56%RegulationElon Musk has lost his lawsuit against Sam Altman and OpenAIIn a significant legal development, Elon Musk's lawsuit against OpenAI co-founders Sam Altman and Greg Brockman has been dismissed by a California jury. The jurTechCrunch AIMay 18· 52%ResearchImport AI 457: AI stuxnet; cursed Muon optimizer; and positive alignmentRecent developments in AI have highlighted both security concerns and advancements in optimization techniques. Researchers have unearthed a 20-year-old virus, fImport AIMay 18· 68%OtherSkillSmith: Compiling Agent Skills into Boundary-Guided Runtime InterfacesThe introduction of SkillSmith marks a significant advancement in the optimization of agent skills for large language model-based systems. This new framework foarXiv cs.AIMay 18· 65%OtherAgentStop: Terminating Local AI Agents Early to Save Energy in Consumer DevicesRecent research has introduced AgentStop, a novel lightweight efficiency supervisor designed to enhance energy efficiency in AI agents that utilize large languaarXiv cs.LGMay 18· 64%SafetyQuantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision LevelsRecent research has highlighted the unintended consequences of post-training quantization on large language models (LLMs), specifically focusing on how this proarXiv cs.LGMay 18· 64%SafetyReducing the Safety Tax in LLM Safety Alignment with On-Policy Self-DistillationRecent advancements in large language model (LLM) safety alignment have been highlighted in a new paper introducing On-Policy Self-Distillation (OPSA). This innarXiv cs.LGMay 18· 63%SafetyFair outputs, Biased Internals: Causal Potency and Asymmetry of Latent Bias in LLMs for High-Stakes DecisionsA recent study has shed light on the internal biases present in instruction-tuned language models (LLMs) that are increasingly used for high-stakes decision-makarXiv cs.AIMay 18· 63%OtherDeepSlide: From Artifacts to Presentation DeliveryDeepSlide is an innovative multi-agent system designed to enhance the process of presentation delivery. By integrating narrative planning, slide generation, andarXiv cs.AIMay 18· 63%OtherCAX-Agent: A Lightweight Agent Harness for Reliable APDL AutomationThe recent paper on CAX-Agent introduces a novel lightweight agent harness aimed at enhancing the reliability of MAPDL finite-element simulations. By incorporatarXiv cs.AIMay 18· 63%InfrastructureGQLA: Group-Query Latent Attention for Hardware-Adaptive Large Language Model DecodingThe recent introduction of Group-Query Latent Attention (GQLA) marks a significant advancement in the field of large language model decoding. This innovative aparXiv cs.LGMay 18· 63%SafetySDOF: Taming the Alignment Tax in Multi-Agent Orchestration with State-Constrained DispatchRecent research has unveiled SDOF, a novel framework aimed at improving multi-agent orchestration by addressing the challenges associated with alignment tax. TharXiv cs.AIMay 18· 62%FundingMuteBench: Modality Unavailability Tolerance Evaluation for Incomplete Multimodal FusionMuteBench has been introduced as a novel benchmark aimed at assessing the performance of multimodal fusion architectures when faced with incomplete data scenariarXiv cs.LGMay 18· 62%OtherICRL: Learning to Internalize Self-Critique with Reinforcement LearningRecent advancements in AI research have led to the introduction of ICRL, a novel framework designed to enhance the capabilities of large language models througharXiv cs.AIMay 18· 62%OtherMask-Morph Graph U-Net: A Generalisable Mesh-Based Surrogate for Crashworthiness Field Prediction under Large Geometric VariationThe introduction of the Mask-Morph Graph U-Net (MMGUNet) marks a significant advancement in the field of crashworthiness field prediction. This innovative machiarXiv cs.LGMay 18· 62%OtherTeamTR: Trust-Region Fine-Tuning for Multi-Agent LLM CoordinationRecent research has unveiled TeamTR, a novel framework designed to improve coordination among multi-agent large language models (LLMs). By employing a trust-regarXiv cs.LGMay 18· 62%OtherNIMO Controller: a self-driving laboratory orchestrator based on the Model Context ProtocolThe introduction of the NIMO Controller marks a significant advancement in the realm of scientific experimentation by utilizing the Model Context Protocol (MCP)arXiv cs.AIMay 18· 62%FundingDoes Theory of Mind Improvement Really Benefit Human-AI Interactions? Empirical Findings from Interactive EvaluationsRecent research has explored the impact of enhancing Theory of Mind (ToM) capabilities in Large Language Models (LLMs) on human-AI interactions. The study introarXiv cs.AIMay 18· 62%OtherReading the Cell, Designing the Cure: Perturbation-Conditioned Molecular Diffusion for Function-Oriented Drug DesignRecent advancements in drug design have been highlighted through the introduction of a novel approach known as Transcriptome-based Drug Design (TBDD), which utiarXiv cs.LGMay 18· 62%OtherLogical Grammar Induction via Graph Kolmogorov Complexity: A Neuro-Symbolic Framework for Self-Healing Clinical Data IntegrityRecent advancements in healthcare data integrity have been highlighted through the introduction of Logic-GNN, a neuro-symbolic framework that leverages Graph KoarXiv cs.LGMay 18· 61%FundingNOVA: Fundamental Limits of Knowledge Discovery Through AIThe NOVA framework has been introduced by researchers to address the fundamental limits of knowledge discovery in AI systems. This framework not only models thearXiv cs.AIMay 18· 61%SafetyVerifiable Agentic Infrastructure: Proof-Derived Authorization for Sovereign AI SystemsRecent research has unveiled a new framework called the Distributed Trust Framework (DTF), which aims to address the risks associated with autonomous AI agents arXiv cs.AIMay 18· 60%FundingPrivacy Evaluation of Generative Models for Trajectory GenerationRecent research has highlighted serious privacy vulnerabilities in generative models designed for trajectory data generation. Despite these models being developarXiv cs.LGMay 18· 60%