https://aiexpert.news/en/article/river-reinforcement-learning-without-ground-truth-enables-cost-effective-llm-fin ai|expert en 2026-06-26T04:28:42.590Z RiVER Enables Reinforcement Learning Without Ground-Truth Labels https://aiexpert.news/pt/article/river-reinforcement-learning-without-ground-truth-enables-cost-effective-llm-fin ai|expert pt-BR 2026-06-26T04:28:42.590Z RiVER Permite Aprendizado por Reforço Sem Rótulos de Verdade Absoluta https://aiexpert.news/es/article/river-reinforcement-learning-without-ground-truth-enables-cost-effective-llm-fin ai|expert es 2026-06-26T04:28:42.590Z RiVER Permite Aprendizaje por Refuerzo Sin Etiquetas de Verdad Absoluta https://aiexpert.news/en/article/world-model-hallucinations-are-predictable-hansen-wangs-mmbench2-fixes-a-product ai|expert en 2026-06-26T04:14:42.554Z World Model Hallucination Is a Data Problem, Not Architecture https://aiexpert.news/pt/article/world-model-hallucinations-are-predictable-hansen-wangs-mmbench2-fixes-a-product ai|expert pt-BR 2026-06-26T04:14:42.554Z A Alucinação de Modelos Mundiais é um Problema de Dados, Não de Arquitetura https://aiexpert.news/es/article/world-model-hallucinations-are-predictable-hansen-wangs-mmbench2-fixes-a-product ai|expert es 2026-06-26T04:14:42.554Z La Alucinación de Modelos Mundiales es un Problema de Datos, No de Arquitectura https://aiexpert.news/en/article/learning-to-fold-vision-language-action-policies-win-icra-robot-challenge ai|expert en 2026-06-26T02:48:42.476Z Single Researcher Places 2nd in ICRA Robot-Folding Challenge https://aiexpert.news/pt/article/learning-to-fold-vision-language-action-policies-win-icra-robot-challenge ai|expert pt-BR 2026-06-26T02:48:42.476Z Pesquisador Solitário Fica em 2º Lugar no Desafio de Robô Dobrando Roupas da ICRA https://aiexpert.news/es/article/learning-to-fold-vision-language-action-policies-win-icra-robot-challenge ai|expert es 2026-06-26T02:48:42.476Z Investigador Solitario Queda en 2º Lugar en el Desafío de Robot Doblador de Ropa de ICRA https://aiexpert.news/en/article/avl-test-data-to-analytics-lakehouse-for-automotive-measurement ai|expert en 2026-06-25T23:52:42.890Z AVL Cuts Test Data Analysis Time From Days to Minutes With Databricks Lakehouse https://aiexpert.news/pt/article/avl-test-data-to-analytics-lakehouse-for-automotive-measurement ai|expert pt-BR 2026-06-25T23:52:42.890Z AVL Reduz Tempo de Análise de Dados de Teste de Dias para Minutos Com Databricks Lakehouse https://aiexpert.news/es/article/avl-test-data-to-analytics-lakehouse-for-automotive-measurement ai|expert es 2026-06-25T23:52:42.890Z AVL Reduce el Tiempo de Análisis de Datos de Prueba de Días a Minutos Con Databricks Lakehouse https://aiexpert.news/en/article/privacy-aware-infrastructure-in-ai-native-era-metas-asset-classification-case-st ai|expert en 2026-06-25T23:20:42.764Z Meta Replaces LLM Calls with Deterministic Rules to Cut Costs https://aiexpert.news/pt/article/privacy-aware-infrastructure-in-ai-native-era-metas-asset-classification-case-st ai|expert pt-BR 2026-06-25T23:20:42.764Z Meta Substitui Chamadas de LLM por Regras Determinísticas para Reduzir Custos https://aiexpert.news/es/article/privacy-aware-infrastructure-in-ai-native-era-metas-asset-classification-case-st ai|expert es 2026-06-25T23:20:42.764Z Meta Reemplaza Llamadas de LLM por Reglas Determinísticas para Reducir Costos https://aiexpert.news/en/article/hugging-face-and-vllm-partner-on-one-command-open-inference ai|expert en 2026-06-25T20:50:42.512Z One Command Spins Up Private vLLM Endpoints at $1.50/Hour https://aiexpert.news/pt/article/hugging-face-and-vllm-partner-on-one-command-open-inference ai|expert pt-BR 2026-06-25T20:50:42.512Z Um Comando Ativa Endpoints Privados vLLM a $1.50/Hora https://aiexpert.news/es/article/hugging-face-and-vllm-partner-on-one-command-open-inference ai|expert es 2026-06-25T20:50:42.512Z Un Comando Activa Endpoints Privados vLLM a $1.50/Hora https://aiexpert.news/en/article/cloudflares-saga-rollback-pattern-distributed-transactions-for-agent-workflows ai|expert en 2026-06-25T14:04:42.674Z Cloudflare Cuts Saga Boilerplate from Agent Workflows https://aiexpert.news/pt/article/cloudflares-saga-rollback-pattern-distributed-transactions-for-agent-workflows ai|expert pt-BR 2026-06-25T14:04:42.674Z Cloudflare Remove Boilerplate de Saga dos Workflows de Agentes https://aiexpert.news/es/article/cloudflares-saga-rollback-pattern-distributed-transactions-for-agent-workflows ai|expert es 2026-06-25T14:04:42.674Z Cloudflare Elimina el Boilerplate de Saga de los Workflows de Agentes https://aiexpert.news/en/article/models-forget-learned-rules-mid-training-understanding-natural-ungrokking ai|expert en 2026-06-25T13:20:42.507Z Models Shed Learned Rules During Training https://aiexpert.news/pt/article/models-forget-learned-rules-mid-training-understanding-natural-ungrokking ai|expert pt-BR 2026-06-25T13:20:42.507Z Modelos Esquecem Regras Aprendidas Durante o Treinamento https://aiexpert.news/es/article/models-forget-learned-rules-mid-training-understanding-natural-ungrokking ai|expert es 2026-06-25T13:20:42.507Z Modelos Olvidan Reglas Aprendidas Durante el Entrenamiento https://aiexpert.news/en/article/process-reward-models-meet-rl-post-training-agents-learn-to-score-their-own-step ai|expert en 2026-06-25T09:48:42.496Z Free Scoring Signal Emerges from Standard RL Post-Training Runs https://aiexpert.news/pt/article/process-reward-models-meet-rl-post-training-agents-learn-to-score-their-own-step ai|expert pt-BR 2026-06-25T09:48:42.496Z Sinal de Pontuação Gratuito Emerge dos Padrões de Pós-Treinamento RL https://aiexpert.news/es/article/process-reward-models-meet-rl-post-training-agents-learn-to-score-their-own-step ai|expert es 2026-06-25T09:48:42.496Z Señal de Puntuación Gratuita Emerge de los Padrón de Pós-Entrenamiento RL https://aiexpert.news/en/article/model-forensics-intent-vs-incompetence-in-ai-behavior ai|expert en 2026-06-25T06:44:42.440Z DeepMind Forensic Protocol Diagnoses Confused vs. Misaligned AI https://aiexpert.news/pt/article/model-forensics-intent-vs-incompetence-in-ai-behavior ai|expert pt-BR 2026-06-25T06:44:42.440Z Protocolo de Perícia do DeepMind Diagnostica IA Confusa vs. Desalinhada https://aiexpert.news/es/article/model-forensics-intent-vs-incompetence-in-ai-behavior ai|expert es 2026-06-25T06:44:42.440Z Protocolo de Análisis Forense de DeepMind Diagnostica IA Confundida vs. Desalineada https://aiexpert.news/en/article/auditing-multimodal-models-for-order-sensitivity-reveals-hidden-reliability-gaps ai|expert en 2026-06-25T05:46:42.770Z Multimodal Models Flip Answers When Evidence Order Changes https://aiexpert.news/pt/article/auditing-multimodal-models-for-order-sensitivity-reveals-hidden-reliability-gaps ai|expert pt-BR 2026-06-25T05:46:42.770Z Modelos Multimodais Trocam Respostas Quando a Ordem de Evidências Muda https://aiexpert.news/es/article/auditing-multimodal-models-for-order-sensitivity-reveals-hidden-reliability-gaps ai|expert es 2026-06-25T05:46:42.770Z Los Modelos Multimodales Cambian de Respuesta Cuando Cambia el Orden de las Evidencias https://aiexpert.news/en/article/production-voice-models-miss-emotional-cuessafety-implications-for-real-time-api ai|expert en 2026-06-25T05:15:52.440Z Production Voice AIs Ignore Emotion, Approving Fraud and Ending Care Calls https://aiexpert.news/pt/article/production-voice-models-miss-emotional-cuessafety-implications-for-real-time-api ai|expert pt-BR 2026-06-25T05:15:52.440Z IAs de Voz em Produção Ignoram Emoção, Aprovando Fraudes e Encerrando Chamadas de Cuidado https://aiexpert.news/es/article/production-voice-models-miss-emotional-cuessafety-implications-for-real-time-api ai|expert es 2026-06-25T05:15:52.440Z Las IAs de Voz en Producción Ignoran la Emoción, Aprobando Fraudes y Terminando Llamadas de Cuidado https://aiexpert.news/en/article/qwen-releases-397b-language-world-models-that-simulate-agent-environments-outper ai|expert en 2026-06-25T04:44:42.798Z Qwen's 397B Model Simulates Agent Environments Better Than GPT-5.4 https://aiexpert.news/pt/article/qwen-releases-397b-language-world-models-that-simulate-agent-environments-outper ai|expert pt-BR 2026-06-25T04:44:42.798Z O Modelo 397B do Qwen Simula Ambientes de Agentes Melhor que o GPT-5.4 https://aiexpert.news/es/article/qwen-releases-397b-language-world-models-that-simulate-agent-environments-outper ai|expert es 2026-06-25T04:44:42.798Z El Modelo 397B de Qwen Simula Entornos de Agentes Mejor que GPT-5.4 https://aiexpert.news/en/article/grab-builds-secure-agentic-ai-runtime-for-payment-and-ops ai|expert en 2026-06-25T04:18:43.687Z Grab Treats Autonomous Agents as Untrusted by Design https://aiexpert.news/pt/article/grab-builds-secure-agentic-ai-runtime-for-payment-and-ops ai|expert pt-BR 2026-06-25T04:18:43.687Z Grab Trata Agentes Autônomos como Não Confiáveis por Design https://aiexpert.news/es/article/grab-builds-secure-agentic-ai-runtime-for-payment-and-ops ai|expert es 2026-06-25T04:18:43.687Z Grab Trata Agentes Autónomos como No Confiables por Diseño https://aiexpert.news/en/article/ffasr-leaderboard-benchmarking-asr-performance-in-real-world-conditions ai|expert en 2026-06-24T19:48:42.518Z FFASR Benchmark Exposes Far-Field Speech Recognition Gap https://aiexpert.news/pt/article/ffasr-leaderboard-benchmarking-asr-performance-in-real-world-conditions ai|expert pt-BR 2026-06-24T19:48:42.518Z Benchmark FFASR Expõe Lacuna no Reconhecimento de Fala em Campo Distante https://aiexpert.news/es/article/ffasr-leaderboard-benchmarking-asr-performance-in-real-world-conditions ai|expert es 2026-06-24T19:48:42.518Z Benchmark FFASR Expone la Brecha en el Reconocimiento de Voz de Campo Lejano https://aiexpert.news/en/article/daikin-builds-consistent-data-pipelines-at-scale-with-databricks-genie ai|expert en 2026-06-24T19:15:54.239Z Daikin Turns Data Pipeline Rules Into Workspace Skills https://aiexpert.news/pt/article/daikin-builds-consistent-data-pipelines-at-scale-with-databricks-genie ai|expert pt-BR 2026-06-24T19:15:54.239Z Daikin Transforma Regras de Pipeline de Dados em Workspace Skills https://aiexpert.news/es/article/daikin-builds-consistent-data-pipelines-at-scale-with-databricks-genie ai|expert es 2026-06-24T19:15:54.239Z Daikin Convierte Reglas de Pipeline de Datos en Workspace Skills https://aiexpert.news/en/article/chinas-cpu-only-supercomputer-topples-us-incumbent-reshapes-chip-strategy ai|expert en 2026-06-24T17:30:42.925Z China's LineShine Tops TOP500 but Lags on AI Training https://aiexpert.news/pt/article/chinas-cpu-only-supercomputer-topples-us-incumbent-reshapes-chip-strategy ai|expert pt-BR 2026-06-24T17:30:42.925Z LineShine da China Lidera TOP500, mas Fica Atrás em Treinamento de IA https://aiexpert.news/es/article/chinas-cpu-only-supercomputer-topples-us-incumbent-reshapes-chip-strategy ai|expert es 2026-06-24T17:30:42.925Z El LineShine de China Lidera el TOP500, pero Se Queda Atrás en Entrenamiento de IA https://aiexpert.news/en/article/nvidia-nemo-automodel-automates-fine-tuning-optimization-cuts-training-latency ai|expert en 2026-06-24T16:18:42.584Z NeMo AutoModel cuts MoE training time by 3.7× on single swap https://aiexpert.news/pt/article/nvidia-nemo-automodel-automates-fine-tuning-optimization-cuts-training-latency ai|expert pt-BR 2026-06-24T16:18:42.584Z NeMo AutoModel reduz tempo de treinamento de MoE em 3.7× com um único swap https://aiexpert.news/es/article/nvidia-nemo-automodel-automates-fine-tuning-optimization-cuts-training-latency ai|expert es 2026-06-24T16:18:42.584Z NeMo AutoModel reduce el tiempo de entrenamiento de MoE en 3.7× con un único swap https://aiexpert.news/en/article/xcures-lands-46m-series-b-for-ai-medical-records-cleanup ai|expert en 2026-06-24T15:42:42.664Z xCures raises $46M to build data-structuring layer for clinical AI https://aiexpert.news/pt/article/xcures-lands-46m-series-b-for-ai-medical-records-cleanup ai|expert pt-BR 2026-06-24T15:42:42.664Z xCures arrecada $46M para construir uma camada de estruturação de dados para IA clínica https://aiexpert.news/es/article/xcures-lands-46m-series-b-for-ai-medical-records-cleanup ai|expert es 2026-06-24T15:42:42.664Z xCures recauda $46M para construir una capa de estructuración de datos para IA clínica https://aiexpert.news/en/article/china-a100-black-market-prices-triple-export-controls-signal-tightens ai|expert en 2026-06-24T10:56:42.596Z A100 GPU Prices Triple in China as Export Gates Slam Shut https://aiexpert.news/pt/article/china-a100-black-market-prices-triple-export-controls-signal-tightens ai|expert pt-BR 2026-06-24T10:56:42.596Z Preços do GPU A100 Triplicam na China com Fechamento das Comportas de Exportação https://aiexpert.news/es/article/china-a100-black-market-prices-triple-export-controls-signal-tightens ai|expert es 2026-06-24T10:56:42.596Z Los Precios del GPU A100 se Triplican en China mientras se Cierran las Compuertas de Exportación https://aiexpert.news/en/article/why-agent-evaluation-breaks-grading-multi-modal-agentic-outputs ai|expert en 2026-06-24T08:15:53.403Z Strict Regex Fix Raises Agent Grading Recall by 60 Percentage Points https://aiexpert.news/pt/article/why-agent-evaluation-breaks-grading-multi-modal-agentic-outputs ai|expert pt-BR 2026-06-24T08:15:53.403Z Correção de Regex Rigoroso Aumenta Recall de Avaliação de Agentes em 60 Pontos Percentuais https://aiexpert.news/es/article/why-agent-evaluation-breaks-grading-multi-modal-agentic-outputs ai|expert es 2026-06-24T08:15:53.403Z Corrección de Regex Estricta Aumenta Recuperación de Calificación de Agentes en 60 Puntos Porcentuales