https://aiexpert.news/en/article/multimodal-model-diffing-identifying-and-controlling-visual-features-in-mllms ai|expert en 2026-08-11T06:36:37.400Z MMDiff Lets Engineers Audit and Edit Multimodal Model Features https://aiexpert.news/pt/article/multimodal-model-diffing-identifying-and-controlling-visual-features-in-mllms ai|expert pt-BR 2026-08-11T06:36:37.400Z MMDiff Permite que Engenheiros Auditem e Editem Características de Modelos Multimodais https://aiexpert.news/es/article/multimodal-model-diffing-identifying-and-controlling-visual-features-in-mllms ai|expert es 2026-08-11T06:36:37.400Z MMDiff Permite que Ingenieros Auditen y Editen Características de Modelos Multimodales https://aiexpert.news/en/article/archagent-v2-using-ai-to-discover-cpu-microarchitecture-in-competition ai|expert en 2026-08-11T06:04:37.483Z ArchAgent v2 beats human champions on three-level cache design https://aiexpert.news/pt/article/archagent-v2-using-ai-to-discover-cpu-microarchitecture-in-competition ai|expert pt-BR 2026-08-11T06:04:37.483Z ArchAgent v2 supera campeões humanos no design de cache de três níveis https://aiexpert.news/es/article/archagent-v2-using-ai-to-discover-cpu-microarchitecture-in-competition ai|expert es 2026-08-11T06:04:37.483Z ArchAgent v2 supera a campeones humanos en diseño de caché de tres niveles https://aiexpert.news/en/article/genco-neural-solvers-for-power-grid-analysiswhen-foundation-models-meet-engineer ai|expert en 2026-08-11T05:32:37.674Z GENCO replaces three classical power grid solvers with one neural architecture https://aiexpert.news/pt/article/genco-neural-solvers-for-power-grid-analysiswhen-foundation-models-meet-engineer ai|expert pt-BR 2026-08-11T05:32:37.674Z GENCO substitui três solucionadores clássicos de rede elétrica por uma arquitetura neural https://aiexpert.news/es/article/genco-neural-solvers-for-power-grid-analysiswhen-foundation-models-meet-engineer ai|expert es 2026-08-11T05:32:37.674Z GENCO reemplaza tres solucionadores clásicos de red eléctrica con una arquitectura neural https://aiexpert.news/en/article/consilience-for-verifier-free-test-time-scaling-confidence-based-reasoning-on-bu ai|expert en 2026-08-11T05:01:34.007Z Amazon's Consilience Detects Silent Mode Collapse in Confidence-Based Reasoning https://aiexpert.news/pt/article/consilience-for-verifier-free-test-time-scaling-confidence-based-reasoning-on-bu ai|expert pt-BR 2026-08-11T05:01:34.007Z Consilience da Amazon Detecta Colapso de Modo Silencioso no Raciocínio Baseado em Confiança https://aiexpert.news/es/article/consilience-for-verifier-free-test-time-scaling-confidence-based-reasoning-on-bu ai|expert es 2026-08-11T05:01:34.007Z Consilience de Amazon Detecta el Colapso del Modo Silencioso en el Razonamiento Basado en Confianza https://aiexpert.news/en/article/decoding-level-taboo-testing-llm-robustness-under-real-deployment-constraints ai|expert en 2026-08-11T04:28:38.420Z Taboo Stress-Tests LLMs on Production Constraints https://aiexpert.news/pt/article/decoding-level-taboo-testing-llm-robustness-under-real-deployment-constraints ai|expert pt-BR 2026-08-11T04:28:38.420Z Taboo Realiza Testes de Estresse de LLMs em Restrições de Produção https://aiexpert.news/es/article/decoding-level-taboo-testing-llm-robustness-under-real-deployment-constraints ai|expert es 2026-08-11T04:28:38.420Z Taboo Realiza Pruebas de Estrés de LLMs en Restricciones de Producción https://aiexpert.news/en/article/dutch-municipalities-adopt-first-values-driven-llm-evaluation-for-public-sector- ai|expert en 2026-08-11T03:58:37.323Z Amsterdam Splits Factuality and Honesty in Language Model Evaluation https://aiexpert.news/pt/article/dutch-municipalities-adopt-first-values-driven-llm-evaluation-for-public-sector- ai|expert pt-BR 2026-08-11T03:58:37.323Z Amsterdam Separa Precisão e Honestidade na Avaliação de Modelos de Linguagem https://aiexpert.news/es/article/dutch-municipalities-adopt-first-values-driven-llm-evaluation-for-public-sector- ai|expert es 2026-08-11T03:58:37.323Z Amsterdam Separa Precisión y Honestidad en la Evaluación de Modelos de Lenguaje https://aiexpert.news/en/article/grounding-enterprise-agents-in-mixed-data-without-losing-governance-a-databricks ai|expert en 2026-08-11T01:15:23.630Z Unity Catalog Routes Agent Security to Data Layer, Not Prompts https://aiexpert.news/pt/article/grounding-enterprise-agents-in-mixed-data-without-losing-governance-a-databricks ai|expert pt-BR 2026-08-11T01:15:23.630Z Unity Catalog Roteia Segurança de Agentes para Camada de Dados, Não para Prompts https://aiexpert.news/es/article/grounding-enterprise-agents-in-mixed-data-without-losing-governance-a-databricks ai|expert es 2026-08-11T01:15:23.630Z Unity Catalog Enruta la Seguridad de Agentes a la Capa de Datos, No a los Prompts https://aiexpert.news/en/article/github-copilot-sdk-for-java-unlocks-enterprise-ide-integration-at-scale ai|expert en 2026-08-10T21:50:37.419Z GitHub Copilot SDK for Java cuts vendor lock-in for enterprise deployments https://aiexpert.news/pt/article/github-copilot-sdk-for-java-unlocks-enterprise-ide-integration-at-scale ai|expert pt-BR 2026-08-10T21:50:37.419Z SDK do GitHub Copilot para Java reduz vendor lock-in em implementações corporativas https://aiexpert.news/es/article/github-copilot-sdk-for-java-unlocks-enterprise-ide-integration-at-scale ai|expert es 2026-08-10T21:50:37.419Z SDK de GitHub Copilot para Java reduce vendor lock-in en implementaciones empresariales https://aiexpert.news/en/article/databricks-omnigent-compositional-safety-policies-for-inference ai|expert en 2026-08-10T18:54:37.409Z Databricks Blocks Prompt-Injection Attacks via Session-State Policy https://aiexpert.news/pt/article/databricks-omnigent-compositional-safety-policies-for-inference ai|expert pt-BR 2026-08-10T18:54:37.409Z Databricks Bloqueia Ataques de Injeção de Prompt via Política de Estado da Sessão https://aiexpert.news/es/article/databricks-omnigent-compositional-safety-policies-for-inference ai|expert es 2026-08-10T18:54:37.409Z Databricks Bloquea Ataques de Inyección de Prompt mediante Política de Estado de Sesión https://aiexpert.news/en/article/us-startup-fields-quantum-sensors-to-reduce-gps-relianceinfrastructure-redundanc ai|expert en 2026-08-10T18:22:37.356Z Dirac Labs Brings Quantum Navigation to GPS-Denied Warfare https://aiexpert.news/pt/article/us-startup-fields-quantum-sensors-to-reduce-gps-relianceinfrastructure-redundanc ai|expert pt-BR 2026-08-10T18:22:37.356Z Dirac Labs Traz Navegação Quântica para Guerras Sem GPS https://aiexpert.news/es/article/us-startup-fields-quantum-sensors-to-reduce-gps-relianceinfrastructure-redundanc ai|expert es 2026-08-10T18:22:37.356Z Dirac Labs Trae Navegación Cuántica a la Guerra Sin GPS https://aiexpert.news/en/article/post-grokking-instability-in-muon-optimized-transformerstraining-collapse-at-rep ai|expert en 2026-08-10T17:50:37.346Z Muon Optimizer Collapse After Grokking Threatens Production Training https://aiexpert.news/pt/article/post-grokking-instability-in-muon-optimized-transformerstraining-collapse-at-rep ai|expert pt-BR 2026-08-10T17:50:37.346Z Colapso do Otimizador Muon Após Grokking Ameaça Treinamento em Produção https://aiexpert.news/es/article/post-grokking-instability-in-muon-optimized-transformerstraining-collapse-at-rep ai|expert es 2026-08-10T17:50:37.346Z Colapso del Optimizador Muon Después de Grokking Amenaza el Entrenamiento en Producción https://aiexpert.news/en/article/open-weights-voice-agents-with-full-deployment-control-nvidia-magpie-tts-at-16-l ai|expert en 2026-08-10T16:50:37.324Z NVIDIA's Magpie TTS Hits 32ms Latency With Open Weights https://aiexpert.news/pt/article/open-weights-voice-agents-with-full-deployment-control-nvidia-magpie-tts-at-16-l ai|expert pt-BR 2026-08-10T16:50:37.324Z O Magpie TTS da NVIDIA Atinge 32ms de Latência Com Pesos Abertos https://aiexpert.news/es/article/open-weights-voice-agents-with-full-deployment-control-nvidia-magpie-tts-at-16-l ai|expert es 2026-08-10T16:50:37.324Z El Magpie TTS de NVIDIA Alcanza 32ms de Latencia Con Pesos Abiertos https://aiexpert.news/en/article/sabre-automated-vlm-stress-testing-catches-failure-modes-pre-deployment ai|expert en 2026-08-10T12:20:37.523Z SABRE Benchmark Exposes Vision Models' Blindness to Visual Contradictions https://aiexpert.news/pt/article/sabre-automated-vlm-stress-testing-catches-failure-modes-pre-deployment ai|expert pt-BR 2026-08-10T12:20:37.523Z Benchmark SABRE Expõe a Cegueira de Modelos de Visão para Contradições Visuais https://aiexpert.news/es/article/sabre-automated-vlm-stress-testing-catches-failure-modes-pre-deployment ai|expert es 2026-08-10T12:20:37.523Z Benchmark SABRE Expone la Ceguera de los Modelos de Visión ante Contradicciones Visuales https://aiexpert.news/en/article/making-knowledge-distillation-cheap-enough-to-run-in-production ai|expert en 2026-08-10T10:46:37.246Z Multiverse Computing Cuts Knowledge Distillation to Single GPU https://aiexpert.news/pt/article/making-knowledge-distillation-cheap-enough-to-run-in-production ai|expert pt-BR 2026-08-10T10:46:37.246Z Multiverse Computing Reduz Knowledge Distillation para Single GPU https://aiexpert.news/es/article/making-knowledge-distillation-cheap-enough-to-run-in-production ai|expert es 2026-08-10T10:46:37.246Z Multiverse Computing Reduce Knowledge Distillation a Single GPU https://aiexpert.news/en/article/post-training-breaks-creativityheres-why-instruction-tuning-might-save-your-reas ai|expert en 2026-08-10T04:00:37.649Z CreativeInstruct Recovers Diversity Without Slowing Inference https://aiexpert.news/pt/article/post-training-breaks-creativityheres-why-instruction-tuning-might-save-your-reas ai|expert pt-BR 2026-08-10T04:00:37.649Z CreativeInstruct Recupera Diversidade Sem Desacelerar a Inferência https://aiexpert.news/es/article/post-training-breaks-creativityheres-why-instruction-tuning-might-save-your-reas ai|expert es 2026-08-10T04:00:37.649Z CreativeInstruct Recupera Diversidad sin Ralentizar la Inferencia https://aiexpert.news/en/article/blast-radius-predictive-memory-management-for-cost-effective-agentic-coding ai|expert en 2026-08-10T03:30:24.236Z Mankind Labs Cuts Agentic Loop Tokens 17–26% With Reversible Memory Eviction https://aiexpert.news/pt/article/blast-radius-predictive-memory-management-for-cost-effective-agentic-coding ai|expert pt-BR 2026-08-10T03:30:24.236Z Mankind Labs Reduz Tokens de Loop Agentic em 17–26% Com Evicção Reversível de Memória https://aiexpert.news/es/article/blast-radius-predictive-memory-management-for-cost-effective-agentic-coding ai|expert es 2026-08-10T03:30:24.236Z Mankind Labs Reduce Tokens de Loop Agentic 17–26% Con Evicción Reversible de Memoria https://aiexpert.news/en/article/coinrag-fine-grained-kv-cache-reuse-cuts-token-cost-in-long-context-rag ai|expert en 2026-08-10T02:58:37.292Z CoinRAG Achieves 5.3% F1 Gain on Multi-Hop RAG with Nugget Caching https://aiexpert.news/pt/article/coinrag-fine-grained-kv-cache-reuse-cuts-token-cost-in-long-context-rag ai|expert pt-BR 2026-08-10T02:58:37.292Z CoinRAG Alcança Ganho de 5,3% F1 em RAG Multi-Salto com Cache de Nuggets de Informação https://aiexpert.news/es/article/coinrag-fine-grained-kv-cache-reuse-cuts-token-cost-in-long-context-rag ai|expert es 2026-08-10T02:58:37.292Z CoinRAG Logra una Ganancia de 5,3% F1 en RAG Multi-Salto con Cache de Pepitas de Información