https://24-ai.news/hr/news/2026-05-05/anthropic-enterprise-services-blackstone-goldman/ 24 AI hr 2026-05-05T00:00:00Z Anthropic s Blackstoneom, Hellman & Friedmanom i Goldman Sachsom osniva enterprise AI uslužnu tvrtku za mid-market https://24-ai.news/hr/news/2026-05-05/arxiv-agentfloor-small-models-tools/ 24 AI hr 2026-05-05T00:00:00Z ArXiv AgentFloor: mali open-weight modeli (0,27B-32B) zadovoljavaju kratkoročne agentne zadatke, GPT-5 zadržava prednost samo u dugoročnom planiranju https://24-ai.news/hr/news/2026-05-05/arxiv-gui-sd-on-policy-self-distillation/ 24 AI hr 2026-05-05T00:00:00Z ArXiv GUI-SD: prvi on-policy self-distillation framework za GUI grounding nadmašuje GRPO na šest benchmarkova u točnosti i efikasnosti treniranja https://24-ai.news/hr/news/2026-05-05/arxiv-saga-gpu-scheduling-agents/ 24 AI hr 2026-05-05T00:00:00Z ArXiv SAGA: workflow-atomic GPU scheduling za AI agente postiže 1,64× brže task completion na 64-GPU klasteru, prihvaćeno na HPDC 2026 https://24-ai.news/hr/news/2026-05-05/arxiv-token-arena-energy-benchmark/ 24 AI hr 2026-05-05T00:00:00Z ArXiv Token Arena: kontinuirani benchmark koji ujedinjuje energiju i kogniciju, otkriva 6,2× razliku u jouleima po točnom odgovoru između endpointa https://24-ai.news/hr/news/2026-05-05/arxiv-vlm-visual-jailbreak-icml-2026/ 24 AI hr 2026-05-05T00:00:00Z ArXiv: Vizualne slike zaobilaze sigurnosne filtre vision-language modela u 40,9 % slučajeva, otkrivaju autori na ICML 2026 https://24-ai.news/hr/news/2026-05-05/aws-agentcore-optimization-preview/ 24 AI hr 2026-05-05T00:00:00Z AWS Bedrock AgentCore Optimization u previewu: automatizirana petlja od produkcijskih traga do A/B testa s OpenTelemetry trace-ovima https://24-ai.news/hr/news/2026-05-05/cncf-github-actions-ci-security/ 24 AI hr 2026-05-05T00:00:00Z CNCF: pinning na immutable digest, least-privilege tokeni i ephemeral runneri — recipe card za sigurniji GitHub Actions pipeline https://24-ai.news/hr/news/2026-05-05/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI hr 2026-05-05T00:00:00Z NIST CAISI: DeepSeek V4 Pro je najsposobniji kineski AI model do sada, ali zaostaje 8 mjeseci za američkim frontierom https://24-ai.news/en/news/2026-05-05/anthropic-enterprise-services-blackstone-goldman/ 24 AI en 2026-05-05T00:00:00Z Anthropic launches enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs for mid-market https://24-ai.news/en/news/2026-05-05/arxiv-agentfloor-small-models-tools/ 24 AI en 2026-05-05T00:00:00Z ArXiv AgentFloor: small open-weight models (0.27B–32B) are sufficient for short-horizon agent tasks; GPT-5 retains advantage only in long-horizon planning https://24-ai.news/en/news/2026-05-05/arxiv-gui-sd-on-policy-self-distillation/ 24 AI en 2026-05-05T00:00:00Z ArXiv GUI-SD: first on-policy self-distillation framework for GUI grounding outperforms GRPO across six benchmarks in accuracy and training efficiency https://24-ai.news/en/news/2026-05-05/arxiv-saga-gpu-scheduling-agents/ 24 AI en 2026-05-05T00:00:00Z ArXiv SAGA: workflow-atomic GPU scheduling for AI agents achieves 1.64× faster task completion on a 64-GPU cluster, accepted at HPDC 2026 https://24-ai.news/en/news/2026-05-05/arxiv-token-arena-energy-benchmark/ 24 AI en 2026-05-05T00:00:00Z ArXiv Token Arena: continuous benchmark unifying energy and cognition reveals 6.2× difference in joules per correct answer across endpoints https://24-ai.news/en/news/2026-05-05/arxiv-vlm-visual-jailbreak-icml-2026/ 24 AI en 2026-05-05T00:00:00Z ArXiv: Visual inputs bypass safety filters in vision-language models 40.9% of the time, ICML 2026 authors find https://24-ai.news/en/news/2026-05-05/aws-agentcore-optimization-preview/ 24 AI en 2026-05-05T00:00:00Z AWS Bedrock AgentCore Optimization in preview: automated loop from production traces to A/B tests via OpenTelemetry https://24-ai.news/en/news/2026-05-05/cncf-github-actions-ci-security/ 24 AI en 2026-05-05T00:00:00Z CNCF: immutable digest pinning, least-privilege tokens, and ephemeral runners — a recipe card for a more secure GitHub Actions pipeline https://24-ai.news/en/news/2026-05-05/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI en 2026-05-05T00:00:00Z NIST CAISI: DeepSeek V4 Pro is the most capable Chinese AI model to date, but trails US frontier by 8 months https://24-ai.news/de/news/2026-05-05/anthropic-enterprise-services-blackstone-goldman/ 24 AI de 2026-05-05T00:00:00Z Anthropic gründet Enterprise-KI-Dienstleistungsunternehmen mit Blackstone, Hellman & Friedman und Goldman Sachs für den Mid-Market https://24-ai.news/de/news/2026-05-05/arxiv-agentfloor-small-models-tools/ 24 AI de 2026-05-05T00:00:00Z ArXiv AgentFloor: Kleine Open-Weight-Modelle (0,27B–32B) reichen für kurzfristige Agenten-Aufgaben aus; GPT-5 behält Vorteil nur bei langfristiger Planung https://24-ai.news/de/news/2026-05-05/arxiv-gui-sd-on-policy-self-distillation/ 24 AI de 2026-05-05T00:00:00Z ArXiv GUI-SD: Erstes On-Policy-Self-Distillation-Framework für GUI-Grounding übertrifft GRPO auf sechs Benchmarks in Genauigkeit und Trainingseffizienz https://24-ai.news/de/news/2026-05-05/arxiv-saga-gpu-scheduling-agents/ 24 AI de 2026-05-05T00:00:00Z ArXiv SAGA: Workflow-atomares GPU-Scheduling für KI-Agenten erreicht 1,64× schnellere Task-Completion auf 64-GPU-Cluster, angenommen auf HPDC 2026 https://24-ai.news/de/news/2026-05-05/arxiv-token-arena-energy-benchmark/ 24 AI de 2026-05-05T00:00:00Z ArXiv Token Arena: kontinuierlicher Benchmark für Energie und Kognition zeigt 6,2-fachen Unterschied in Joule pro korrekter Antwort zwischen Endpunkten https://24-ai.news/de/news/2026-05-05/arxiv-vlm-visual-jailbreak-icml-2026/ 24 AI de 2026-05-05T00:00:00Z ArXiv: Visuelle Eingaben umgehen Sicherheitsfilter von Vision-Language-Modellen in 40,9 % der Fälle, zeigt ICML-2026-Studie https://24-ai.news/de/news/2026-05-05/aws-agentcore-optimization-preview/ 24 AI de 2026-05-05T00:00:00Z AWS Bedrock AgentCore Optimization in der Vorschau: automatisierte Schleife von Produktions-Traces bis A/B-Tests via OpenTelemetry https://24-ai.news/de/news/2026-05-05/cncf-github-actions-ci-security/ 24 AI de 2026-05-05T00:00:00Z CNCF: Unveränderliches Digest-Pinning, Least-Privilege-Token und ephemere Runner — Rezeptkarte für sicherere GitHub-Actions-Pipelines https://24-ai.news/de/news/2026-05-05/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI de 2026-05-05T00:00:00Z NIST CAISI: DeepSeek V4 Pro ist bisher fähigstes chinesisches KI-Modell, liegt aber 8 Monate hinter US-Frontier https://24-ai.news/zh/news/2026-05-05/anthropic-enterprise-services-blackstone-goldman/ 24 AI zh 2026-05-05T00:00:00Z Anthropic联合Blackstone、Hellman & Friedman和Goldman Sachs成立面向中端市场的企业AI服务公司 https://24-ai.news/zh/news/2026-05-05/arxiv-agentfloor-small-models-tools/ 24 AI zh 2026-05-05T00:00:00Z ArXiv AgentFloor：小型开放权重模型（0.27B-32B）能胜任短期智能体任务，GPT-5仅在长期规划上保持优势 https://24-ai.news/zh/news/2026-05-05/arxiv-gui-sd-on-policy-self-distillation/ 24 AI zh 2026-05-05T00:00:00Z ArXiv GUI-SD：首个面向GUI定位的在线自蒸馏框架，在六个基准上超越GRPO强化学习 https://24-ai.news/zh/news/2026-05-05/arxiv-saga-gpu-scheduling-agents/ 24 AI zh 2026-05-05T00:00:00Z ArXiv SAGA：AI智能体的工作流原子化GPU调度在64-GPU集群上实现1.64倍任务完成提速，被HPDC 2026接收 https://24-ai.news/zh/news/2026-05-05/arxiv-token-arena-energy-benchmark/ 24 AI zh 2026-05-05T00:00:00Z ArXiv Token Arena：统一能耗与认知的持续基准，揭示端点间每正确答案能耗6.2倍差距 https://24-ai.news/zh/news/2026-05-05/arxiv-vlm-visual-jailbreak-icml-2026/ 24 AI zh 2026-05-05T00:00:00Z ArXiv：视觉图像以40.9%的成功率绕过视觉语言模型安全过滤器，ICML 2026论文揭示 https://24-ai.news/zh/news/2026-05-05/aws-agentcore-optimization-preview/ 24 AI zh 2026-05-05T00:00:00Z AWS Bedrock AgentCore Optimization进入预览：从生产追踪到A/B测试的自动化循环，基于OpenTelemetry追踪 https://24-ai.news/zh/news/2026-05-05/cncf-github-actions-ci-security/ 24 AI zh 2026-05-05T00:00:00Z CNCF：固定到不可变摘要、最小权限令牌和临时运行器——更安全的GitHub Actions管道实践指南 https://24-ai.news/zh/news/2026-05-05/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI zh 2026-05-05T00:00:00Z NIST CAISI：DeepSeek V4 Pro是迄今最强中国AI模型，但落后美国前沿约8个月 https://24-ai.news/ja/news/2026-05-05/anthropic-enterprise-services-blackstone-goldman/ 24 AI ja 2026-05-05T00:00:00Z AnthropicがBlackstone、Hellman & Friedman、Goldman Sachsと中堅市場向けエンタープライズAIサービス会社を設立 https://24-ai.news/ja/news/2026-05-05/arxiv-agentfloor-small-models-tools/ 24 AI ja 2026-05-05T00:00:00Z ArXiv AgentFloor：小型オープンウェイトモデル（0.27B-32B）が短期エージェントタスクに十分、GPT-5は長期計画のみで優位を維持 https://24-ai.news/ja/news/2026-05-05/arxiv-gui-sd-on-policy-self-distillation/ 24 AI ja 2026-05-05T00:00:00Z ArXiv GUI-SD：GUIグラウンディング向け初のオンポリシー自己蒸留フレームワーク、6つのベンチマークでGRPO強化学習を凌駕 https://24-ai.news/ja/news/2026-05-05/arxiv-saga-gpu-scheduling-agents/ 24 AI ja 2026-05-05T00:00:00Z ArXiv SAGA：AIエージェント向けワークフロー原子化GPUスケジューリング、64-GPUクラスターでタスク完了を1.64倍高速化、HPDC 2026採択 https://24-ai.news/ja/news/2026-05-05/arxiv-token-arena-energy-benchmark/ 24 AI ja 2026-05-05T00:00:00Z ArXiv Token Arena：エネルギーと認知を統合する継続的ベンチマーク、エンドポイント間で正解あたりエネルギーの6.2倍の差を発見 https://24-ai.news/ja/news/2026-05-05/arxiv-vlm-visual-jailbreak-icml-2026/ 24 AI ja 2026-05-05T00:00:00Z ArXiv：視覚画像がVLMの安全フィルターを40.9%の確率で回避、ICML 2026論文が明らかに https://24-ai.news/ja/news/2026-05-05/aws-agentcore-optimization-preview/ 24 AI ja 2026-05-05T00:00:00Z AWS Bedrock AgentCore Optimizationがプレビュー公開：OpenTelemetryトレースで本番環境からA/Bテストまでの自動化ループを実現 https://24-ai.news/ja/news/2026-05-05/cncf-github-actions-ci-security/ 24 AI ja 2026-05-05T00:00:00Z CNCF：不変ダイジェストへのピン留め、最小権限トークン、エフェメラルランナー——より安全なGitHub ActionsパイプラインへのレシピカードCNCF発表 https://24-ai.news/ja/news/2026-05-05/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI ja 2026-05-05T00:00:00Z NIST CAISI：DeepSeek V4 Proはこれまで評価した中で最も優れた中国AIモデルだが、米国フロンティアに8ヶ月遅れ https://24-ai.news/ko/news/2026-05-05/anthropic-enterprise-services-blackstone-goldman/ 24 AI ko 2026-05-05T00:00:00Z Anthropic, Blackstone·Hellman & Friedman·Goldman Sachs와 중소기업 대상 엔터프라이즈 AI 서비스 회사 설립 https://24-ai.news/ko/news/2026-05-05/arxiv-agentfloor-small-models-tools/ 24 AI ko 2026-05-05T00:00:00Z ArXiv AgentFloor：소형 오픈웨이트 모델(0.27B-32B)이 단기 에이전트 작업에 충분, GPT-5는 장기 계획에서만 우위 https://24-ai.news/ko/news/2026-05-05/arxiv-gui-sd-on-policy-self-distillation/ 24 AI ko 2026-05-05T00:00:00Z ArXiv GUI-SD：GUI 그라운딩을 위한 최초의 온폴리시 자기 증류 프레임워크, 6개 벤치마크에서 GRPO 강화학습 능가 https://24-ai.news/ko/news/2026-05-05/arxiv-saga-gpu-scheduling-agents/ 24 AI ko 2026-05-05T00:00:00Z ArXiv SAGA：AI 에이전트를 위한 워크플로우 원자화 GPU 스케줄링, 64-GPU 클러스터에서 작업 완료 1.64배 단축, HPDC 2026 채택 https://24-ai.news/ko/news/2026-05-05/arxiv-token-arena-energy-benchmark/ 24 AI ko 2026-05-05T00:00:00Z ArXiv Token Arena：에너지와 인지를 통합한 지속적 벤치마크, 엔드포인트 간 정답당 에너지 6.2배 차이 발견 https://24-ai.news/ko/news/2026-05-05/arxiv-vlm-visual-jailbreak-icml-2026/ 24 AI ko 2026-05-05T00:00:00Z ArXiv：시각 이미지가 VLM 안전 필터를 40.9% 확률로 우회, ICML 2026 논문 공개 https://24-ai.news/ko/news/2026-05-05/aws-agentcore-optimization-preview/ 24 AI ko 2026-05-05T00:00:00Z AWS Bedrock AgentCore Optimization 프리뷰 출시：OpenTelemetry 트레이스로 생산에서 A/B 테스트까지 자동화 루프 구현 https://24-ai.news/ko/news/2026-05-05/cncf-github-actions-ci-security/ 24 AI ko 2026-05-05T00:00:00Z CNCF：불변 다이제스트 고정, 최소 권한 토큰, 임시 러너——더 안전한 GitHub Actions 파이프라인을 위한 레시피 카드 https://24-ai.news/ko/news/2026-05-05/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI ko 2026-05-05T00:00:00Z NIST CAISI：DeepSeek V4 Pro, 지금까지 평가된 최강 중국 AI 모델이지만 미국 프런티어에 8개월 뒤처져 https://24-ai.news/hr/news/2026-05-04/arxiv-adamezo-zeroth-order-llm-finetuning/ 24 AI hr 2026-05-04T00:00:00Z AdaMeZO: fino ugađanje LLM-ova Adam-stilom bez pohrane momenata u GPU memoriji https://24-ai.news/hr/news/2026-05-04/arxiv-aem-adaptive-entropy-modulation-rl-agenti/ 24 AI hr 2026-05-04T00:00:00Z ArXiv AEM: adaptivna modulacija entropije za multi-turn RL agente postiže +1,4 % na SWE-bench Verified https://24-ai.news/hr/news/2026-05-04/arxiv-armor-2025-vojni-llm-sigurnost/ 24 AI hr 2026-05-04T00:00:00Z ArXiv ARMOR 2025: prvi vojni benchmark za LLM sigurnost s 519 promptova kroz 21 komercijalni model https://24-ai.news/hr/news/2026-05-04/arxiv-bayes-consistent-agentic-orchestration-icml/ 24 AI hr 2026-05-04T00:00:00Z Position paper s 30 autora na ICML 2026: orkestracija agentnih AI sustava mora biti Bayes-konzistentna https://24-ai.news/hr/news/2026-05-04/arxiv-bwla-w1ax-kvantizacija-llm/ 24 AI hr 2026-05-04T00:00:00Z BWLA: 1-bitna kvantizacija LLM-ova s 3,26× ubrzanjem i 70% boljim rezultatima (ACL 2026) https://24-ai.news/hr/news/2026-05-04/arxiv-stable-gflownet-llm-red-teaming-icml/ 24 AI hr 2026-05-04T00:00:00Z ICML 2026 Spotlight: Stable-GFlowNet uvodi stabilnije i raznovrsnije automatizirano red-teamanje LLM-ova https://24-ai.news/hr/news/2026-05-04/arxiv-tool-calling-framework-llm-agenti/ 24 AI hr 2026-05-04T00:00:00Z ArXiv okvir 'To Call or Not to Call' otkriva da LLM-ovi pogrešno procjenjuju kad im trebaju vanjski alati https://24-ai.news/hr/news/2026-05-04/arxiv-tool-use-tax-llm-agenti/ 24 AI hr 2026-05-04T00:00:00Z ArXiv: skriveni trošak alata u LLM agentima — "tool-use tax" smanjuje točnost čak i kad alati pomažu https://24-ai.news/hr/news/2026-05-04/ibm-ceo-study-c-suite-ai-restrukturiranje/ 24 AI hr 2026-05-04T00:00:00Z IBM studija: 76 % organizacija ima Chief AI Officera, CEO-i očekuju 48 % autonomnih AI odluka do 2030. https://24-ai.news/en/news/2026-05-04/arxiv-adamezo-zeroth-order-llm-finetuning/ 24 AI en 2026-05-04T00:00:00Z AdaMeZO: Adam-style LLM fine-tuning without storing gradient moments in GPU memory https://24-ai.news/en/news/2026-05-04/arxiv-aem-adaptive-entropy-modulation-rl-agenti/ 24 AI en 2026-05-04T00:00:00Z ArXiv AEM: Adaptive Entropy Modulation for multi-turn RL agents achieves +1.4% on SWE-bench Verified https://24-ai.news/en/news/2026-05-04/arxiv-armor-2025-vojni-llm-sigurnost/ 24 AI en 2026-05-04T00:00:00Z ArXiv ARMOR 2025: first military LLM safety benchmark with 519 prompts across 21 commercial models https://24-ai.news/en/news/2026-05-04/arxiv-bayes-consistent-agentic-orchestration-icml/ 24 AI en 2026-05-04T00:00:00Z Position paper by 30 authors at ICML 2026: agentic AI orchestration must be Bayes-consistent https://24-ai.news/en/news/2026-05-04/arxiv-bwla-w1ax-kvantizacija-llm/ 24 AI en 2026-05-04T00:00:00Z BWLA: 1-bit LLM quantization with 3.26× speedup and 70% better results (ACL 2026) https://24-ai.news/en/news/2026-05-04/arxiv-stable-gflownet-llm-red-teaming-icml/ 24 AI en 2026-05-04T00:00:00Z ICML 2026 Spotlight: Stable-GFlowNet introduces more stable and diverse automated LLM red-teaming https://24-ai.news/en/news/2026-05-04/arxiv-tool-calling-framework-llm-agenti/ 24 AI en 2026-05-04T00:00:00Z ArXiv 'To Call or Not to Call' framework reveals LLMs misjudge when they need external tools https://24-ai.news/en/news/2026-05-04/arxiv-tool-use-tax-llm-agenti/ 24 AI en 2026-05-04T00:00:00Z ArXiv: the hidden cost of tools in LLM agents — 'tool-use tax' reduces accuracy even when tools help https://24-ai.news/en/news/2026-05-04/ibm-ceo-study-c-suite-ai-restrukturiranje/ 24 AI en 2026-05-04T00:00:00Z IBM study: 76% of organizations have a Chief AI Officer, CEOs expect 48% autonomous AI decisions by 2030 https://24-ai.news/de/news/2026-05-04/arxiv-adamezo-zeroth-order-llm-finetuning/ 24 AI de 2026-05-04T00:00:00Z AdaMeZO: Adam-Stil LLM-Fine-Tuning ohne Speicherung von Gradientenmomenten im GPU-Speicher https://24-ai.news/de/news/2026-05-04/arxiv-aem-adaptive-entropy-modulation-rl-agenti/ 24 AI de 2026-05-04T00:00:00Z ArXiv AEM: Adaptive Entropiemodulation für Multi-Turn-RL-Agenten erreicht +1,4 % auf SWE-bench Verified https://24-ai.news/de/news/2026-05-04/arxiv-armor-2025-vojni-llm-sigurnost/ 24 AI de 2026-05-04T00:00:00Z ArXiv ARMOR 2025: erster militärischer LLM-Sicherheitsbenchmark mit 519 Prompts über 21 kommerzielle Modelle https://24-ai.news/de/news/2026-05-04/arxiv-bayes-consistent-agentic-orchestration-icml/ 24 AI de 2026-05-04T00:00:00Z Position Paper von 30 Autoren auf ICML 2026: Orchestrierung agentischer KI-Systeme muss Bayes-konsistent sein https://24-ai.news/de/news/2026-05-04/arxiv-bwla-w1ax-kvantizacija-llm/ 24 AI de 2026-05-04T00:00:00Z BWLA: 1-Bit-Quantisierung von Sprachmodellen mit 3,26-facher Beschleunigung und 70 % besseren Ergebnissen (ACL 2026) https://24-ai.news/de/news/2026-05-04/arxiv-stable-gflownet-llm-red-teaming-icml/ 24 AI de 2026-05-04T00:00:00Z ICML 2026 Spotlight: Stable-GFlowNet führt stabileres und vielfältigeres automatisiertes Red-Teaming von Sprachmodellen ein https://24-ai.news/de/news/2026-05-04/arxiv-tool-calling-framework-llm-agenti/ 24 AI de 2026-05-04T00:00:00Z ArXiv-Rahmen 'To Call or Not to Call' zeigt: Sprachmodelle beurteilen falsch, wann sie externe Werkzeuge brauchen https://24-ai.news/de/news/2026-05-04/arxiv-tool-use-tax-llm-agenti/ 24 AI de 2026-05-04T00:00:00Z ArXiv: die versteckten Kosten von Werkzeugen in LLM-Agenten — 'Tool-Use Tax' senkt Genauigkeit selbst wenn Werkzeuge helfen https://24-ai.news/de/news/2026-05-04/ibm-ceo-study-c-suite-ai-restrukturiranje/ 24 AI de 2026-05-04T00:00:00Z IBM-Studie: 76 % der Unternehmen haben einen Chief AI Officer, CEOs erwarten 48 % autonome KI-Entscheidungen bis 2030 https://24-ai.news/zh/news/2026-05-04/arxiv-adamezo-zeroth-order-llm-finetuning/ 24 AI zh 2026-05-04T00:00:00Z AdaMeZO：以类Adam方式微调LLM，无需在GPU内存中存储动量 https://24-ai.news/zh/news/2026-05-04/arxiv-aem-adaptive-entropy-modulation-rl-agenti/ 24 AI zh 2026-05-04T00:00:00Z ArXiv AEM：多轮RL智能体的自适应熵调制在SWE-bench Verified上提升+1.4% https://24-ai.news/zh/news/2026-05-04/arxiv-armor-2025-vojni-llm-sigurnost/ 24 AI zh 2026-05-04T00:00:00Z ArXiv ARMOR 2025：519个提示词测试21个商业LLM的军事安全性基准 https://24-ai.news/zh/news/2026-05-04/arxiv-bayes-consistent-agentic-orchestration-icml/ 24 AI zh 2026-05-04T00:00:00Z ICML 2026 立场论文：30位作者认为智能体AI系统的编排必须符合贝叶斯一致性 https://24-ai.news/zh/news/2026-05-04/arxiv-bwla-w1ax-kvantizacija-llm/ 24 AI zh 2026-05-04T00:00:00Z BWLA：1位量化LLM实现3.26倍加速和70%更好结果（ACL 2026） https://24-ai.news/zh/news/2026-05-04/arxiv-stable-gflownet-llm-red-teaming-icml/ 24 AI zh 2026-05-04T00:00:00Z ICML 2026 Spotlight：Stable-GFlowNet引入更稳定、更多样化的LLM自动化红队测试 https://24-ai.news/zh/news/2026-05-04/arxiv-tool-calling-framework-llm-agenti/ 24 AI zh 2026-05-04T00:00:00Z ArXiv框架「是否调用」揭示LLM错误判断何时需要外部工具 https://24-ai.news/zh/news/2026-05-04/arxiv-tool-use-tax-llm-agenti/ 24 AI zh 2026-05-04T00:00:00Z ArXiv：LLM智能体工具的隐性成本——「工具使用税」即使工具有帮助也会降低准确性 https://24-ai.news/zh/news/2026-05-04/ibm-ceo-study-c-suite-ai-restrukturiranje/ 24 AI zh 2026-05-04T00:00:00Z IBM研究：76%的组织有首席AI官，CEO预计到2030年48%的运营决策将由AI自主作出 https://24-ai.news/ja/news/2026-05-04/arxiv-adamezo-zeroth-order-llm-finetuning/ 24 AI ja 2026-05-04T00:00:00Z AdaMeZO：GPU メモリにモーメントを保存せずAdam方式でLLMをファインチューニング https://24-ai.news/ja/news/2026-05-04/arxiv-aem-adaptive-entropy-modulation-rl-agenti/ 24 AI ja 2026-05-04T00:00:00Z ArXiv AEM：マルチターンRL エージェントの適応的エントロピー変調がSWE-bench Verifiedで+1.4%を達成 https://24-ai.news/ja/news/2026-05-04/arxiv-armor-2025-vojni-llm-sigurnost/ 24 AI ja 2026-05-04T00:00:00Z ArXiv ARMOR 2025：519の軍事プロンプトで21の商用LLMの安全性を評価する初の軍事ベンチマーク https://24-ai.news/ja/news/2026-05-04/arxiv-bayes-consistent-agentic-orchestration-icml/ 24 AI ja 2026-05-04T00:00:00Z ICML 2026立場論文：30名の著者がエージェントAIのオーケストレーションはベイズ一貫性を持つべきと主張 https://24-ai.news/ja/news/2026-05-04/arxiv-bwla-w1ax-kvantizacija-llm/ 24 AI ja 2026-05-04T00:00:00Z BWLA：1ビット量子化LLMで3.26倍の高速化と70%の改善を達成（ACL 2026） https://24-ai.news/ja/news/2026-05-04/arxiv-stable-gflownet-llm-red-teaming-icml/ 24 AI ja 2026-05-04T00:00:00Z ICML 2026 Spotlight：Stable-GFlowNetがより安定した多様なLLM自動レッドチーミングを実現 https://24-ai.news/ja/news/2026-05-04/arxiv-tool-calling-framework-llm-agenti/ 24 AI ja 2026-05-04T00:00:00Z ArXivフレームワーク「呼ぶべきか否か」がLLMの外部ツール判断ミスを明らかに https://24-ai.news/ja/news/2026-05-04/arxiv-tool-use-tax-llm-agenti/ 24 AI ja 2026-05-04T00:00:00Z ArXiv：LLMエージェントのツールの隠れたコスト——「ツール使用税」はツールが役立つ時でも精度を下げる https://24-ai.news/ja/news/2026-05-04/ibm-ceo-study-c-suite-ai-restrukturiranje/ 24 AI ja 2026-05-04T00:00:00Z IBM調査：76%の組織がChief AI Officerを設置、CEOは2030年までにAIが48%の運営決定を自律的に下すと予測 https://24-ai.news/ko/news/2026-05-04/arxiv-adamezo-zeroth-order-llm-finetuning/ 24 AI ko 2026-05-04T00:00:00Z AdaMeZO: GPU 메모리에 모멘트를 저장하지 않고 Adam 방식으로 LLM 파인튜닝하는 새 최적화기 https://24-ai.news/ko/news/2026-05-04/arxiv-aem-adaptive-entropy-modulation-rl-agenti/ 24 AI ko 2026-05-04T00:00:00Z ArXiv AEM: 멀티턴 RL 에이전트를 위한 적응형 엔트로피 변조, SWE-bench Verified에서 +1.4% 향상 https://24-ai.news/ko/news/2026-05-04/arxiv-armor-2025-vojni-llm-sigurnost/ 24 AI ko 2026-05-04T00:00:00Z ArXiv ARMOR 2025: 519개 프롬프트로 21개 상용 LLM의 군사 안전성을 평가한 최초의 군사 벤치마크 https://24-ai.news/ko/news/2026-05-04/arxiv-bayes-consistent-agentic-orchestration-icml/ 24 AI ko 2026-05-04T00:00:00Z ICML 2026 포지션 페이퍼: 저자 30명, 에이전트 AI 오케스트레이션은 베이즈 일관성을 가져야 한다 https://24-ai.news/ko/news/2026-05-04/arxiv-bwla-w1ax-kvantizacija-llm/ 24 AI ko 2026-05-04T00:00:00Z BWLA: 1비트 양자화 LLM으로 3.26배 가속 및 70% 향상 달성 (ACL 2026) https://24-ai.news/ko/news/2026-05-04/arxiv-stable-gflownet-llm-red-teaming-icml/ 24 AI ko 2026-05-04T00:00:00Z ICML 2026 Spotlight: Stable-GFlowNet, 더 안정적이고 다양한 LLM 자동화 레드팀 테스트 도입 https://24-ai.news/ko/news/2026-05-04/arxiv-tool-calling-framework-llm-agenti/ 24 AI ko 2026-05-04T00:00:00Z ArXiv 프레임워크 '호출할 것인가 말 것인가': LLM이 외부 도구 필요성을 잘못 판단한다는 것을 밝혀 https://24-ai.news/ko/news/2026-05-04/arxiv-tool-use-tax-llm-agenti/ 24 AI ko 2026-05-04T00:00:00Z ArXiv: LLM 에이전트 도구의 숨겨진 비용 - '도구 사용세'는 도구가 도움이 될 때도 정확도를 낮춘다 https://24-ai.news/ko/news/2026-05-04/ibm-ceo-study-c-suite-ai-restrukturiranje/ 24 AI ko 2026-05-04T00:00:00Z IBM 연구: 조직의 76%가 최고AI책임자를 보유, CEO들은 2030년까지 AI가 운영 결정의 48%를 자율적으로 내릴 것으로 예상 https://24-ai.news/hr/news/2026-05-02/arxiv-exploration-hacking-rl-resistance/ 24 AI hr 2026-05-02T00:00:00Z Exploration Hacking: mogu li LLM-ovi naučiti opirati se RL treningu i strategijski potiskivati vlastite sposobnosti? https://24-ai.news/hr/news/2026-05-02/arxiv-kellybench-premier-league-decisions/ 24 AI hr 2026-05-02T00:00:00Z KellyBench: AI agenti upravljali kladioničarskim bankrollom u Premier Ligi — svi vodeći modeli izgubili novac https://24-ai.news/hr/news/2026-05-02/arxiv-latent-grpo-reasoning-rl/ 24 AI hr 2026-05-02T00:00:00Z Latent-GRPO: stabilna RL optimizacija za latent reasoning — 7,86 boda na GSM8K-Aug i 4,27 boda na AIME uz 3-4× kraće reasoning chain-ove https://24-ai.news/hr/news/2026-05-02/arxiv-mcphunt-mcp-credential-propagation/ 24 AI hr 2026-05-02T00:00:00Z MCPHunt: prvi benchmark koji mjeri curenje vjerodajnica između granica povjerenja u multi-server MCP agentima — stope 11,5–41,3 % https://24-ai.news/hr/news/2026-05-02/arxiv-msr-synthetic-computers-1000-scale/ 24 AI hr 2026-05-02T00:00:00Z Microsoft Research Synthetic Computers: 1 000 sintetičkih računala kao supstrat za long-horizon trening produktivnih AI agenata https://24-ai.news/hr/news/2026-05-02/aws-transform-bi-migration-bedrock-agentcore/ 24 AI hr 2026-05-02T00:00:00Z AWS Transform automatizira migraciju BI dashboarda iz Tableau i Power BI u QuickSight za dane umjesto mjeseci https://24-ai.news/hr/news/2026-05-02/github-copilot-gpt-52-deprecation-june-2026/ 24 AI hr 2026-05-02T00:00:00Z GitHub povlači GPT-5.2 i GPT-5.2-Codex iz Copilota 1. lipnja 2026. — migracija na GPT-5.5 i GPT-5.3-Codex https://24-ai.news/hr/news/2026-05-02/google-research-open-science-250k-reach/ 24 AI hr 2026-05-02T00:00:00Z Google Research open-source alati dosegnuli 250.000 istraživača: od genoma do monsunskih prognoza za 38 milijuna farmera https://24-ai.news/hr/news/2026-05-02/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI hr 2026-05-02T00:00:00Z NIST CAISI evaluacija DeepSeek V4 Pro: 8 mjeseci zaostatka za frontier US modelima na 9 benchmarka u 5 domena https://24-ai.news/en/news/2026-05-02/arxiv-exploration-hacking-rl-resistance/ 24 AI en 2026-05-02T00:00:00Z Exploration Hacking: Can LLMs Learn to Resist RL Training and Strategically Suppress Their Own Capabilities? https://24-ai.news/en/news/2026-05-02/arxiv-kellybench-premier-league-decisions/ 24 AI en 2026-05-02T00:00:00Z KellyBench: AI agents managing a betting bankroll through the Premier League season — all leading models lost money https://24-ai.news/en/news/2026-05-02/arxiv-latent-grpo-reasoning-rl/ 24 AI en 2026-05-02T00:00:00Z Latent-GRPO: Stable RL Optimization for Latent Reasoning — 7.86 Points on GSM8K-Aug and 4.27 Points on AIME With 3-4× Shorter Reasoning Chains https://24-ai.news/en/news/2026-05-02/arxiv-mcphunt-mcp-credential-propagation/ 24 AI en 2026-05-02T00:00:00Z MCPHunt: first benchmark measuring credential leakage across trust boundaries in multi-server MCP agents — rates of 11.5–41.3% https://24-ai.news/en/news/2026-05-02/arxiv-msr-synthetic-computers-1000-scale/ 24 AI en 2026-05-02T00:00:00Z Microsoft Research Synthetic Computers: 1,000 synthetic computers as a substrate for long-horizon training of productive AI agents https://24-ai.news/en/news/2026-05-02/aws-transform-bi-migration-bedrock-agentcore/ 24 AI en 2026-05-02T00:00:00Z AWS Transform automates BI dashboard migration from Tableau and Power BI to QuickSight in days instead of months https://24-ai.news/en/news/2026-05-02/github-copilot-gpt-52-deprecation-june-2026/ 24 AI en 2026-05-02T00:00:00Z GitHub is retiring GPT-5.2 and GPT-5.2-Codex from Copilot on June 1, 2026 — migration to GPT-5.5 and GPT-5.3-Codex https://24-ai.news/en/news/2026-05-02/google-research-open-science-250k-reach/ 24 AI en 2026-05-02T00:00:00Z Google Research open-source tools reach 250,000 researchers: from genomes to monsoon forecasts for 38 million farmers https://24-ai.news/en/news/2026-05-02/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI en 2026-05-02T00:00:00Z NIST CAISI evaluation of DeepSeek V4 Pro: 8-month lag behind frontier US models across 9 benchmarks in 5 domains https://24-ai.news/de/news/2026-05-02/arxiv-exploration-hacking-rl-resistance/ 24 AI de 2026-05-02T00:00:00Z Exploration Hacking: Können LLMs lernen, sich dem RL-Training zu widersetzen und ihre eigenen Fähigkeiten strategisch zu unterdrücken? https://24-ai.news/de/news/2026-05-02/arxiv-kellybench-premier-league-decisions/ 24 AI de 2026-05-02T00:00:00Z KellyBench: KI-Agenten verwalten Wett-Bankroll durch die Premier-League-Saison — alle führenden Modelle verloren Geld https://24-ai.news/de/news/2026-05-02/arxiv-latent-grpo-reasoning-rl/ 24 AI de 2026-05-02T00:00:00Z Latent-GRPO: Stabile RL-Optimierung für Latent Reasoning — 7,86 Punkte auf GSM8K-Aug und 4,27 Punkte auf AIME bei 3-4× kürzeren Reasoning-Ketten https://24-ai.news/de/news/2026-05-02/arxiv-mcphunt-mcp-credential-propagation/ 24 AI de 2026-05-02T00:00:00Z MCPHunt: erster Benchmark zur Messung von Credential-Leakage über Vertrauensgrenzen in Multi-Server-MCP-Agenten — Raten von 11,5–41,3 % https://24-ai.news/de/news/2026-05-02/arxiv-msr-synthetic-computers-1000-scale/ 24 AI de 2026-05-02T00:00:00Z Microsoft Research Synthetic Computers: 1.000 synthetische Computer als Substrat für das Long-Horizon-Training produktiver KI-Agenten https://24-ai.news/de/news/2026-05-02/aws-transform-bi-migration-bedrock-agentcore/ 24 AI de 2026-05-02T00:00:00Z AWS Transform automatisiert BI-Dashboard-Migration von Tableau und Power BI nach QuickSight in Tagen statt Monaten https://24-ai.news/de/news/2026-05-02/github-copilot-gpt-52-deprecation-june-2026/ 24 AI de 2026-05-02T00:00:00Z GitHub stellt GPT-5.2 und GPT-5.2-Codex in Copilot am 1. Juni 2026 ein — Migration auf GPT-5.5 und GPT-5.3-Codex https://24-ai.news/de/news/2026-05-02/google-research-open-science-250k-reach/ 24 AI de 2026-05-02T00:00:00Z Google Research Open-Source-Tools erreichen 250.000 Forscher: von Genomen bis zu Monsun-Prognosen für 38 Millionen Landwirte https://24-ai.news/de/news/2026-05-02/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI de 2026-05-02T00:00:00Z NIST CAISI-Evaluierung von DeepSeek V4 Pro: 8 Monate Rückstand gegenüber US-Frontier-Modellen in 9 Benchmarks und 5 Domänen https://24-ai.news/zh/news/2026-05-02/arxiv-exploration-hacking-rl-resistance/ 24 AI zh 2026-05-02T00:00:00Z 探索黑客攻击：大语言模型能否学会抵抗强化学习训练并战略性地压制自身能力？ https://24-ai.news/zh/news/2026-05-02/arxiv-kellybench-premier-league-decisions/ 24 AI zh 2026-05-02T00:00:00Z KellyBench：AI代理管理Premier League赛季投注资金——所有顶级模型均亏损 https://24-ai.news/zh/news/2026-05-02/arxiv-latent-grpo-reasoning-rl/ 24 AI zh 2026-05-02T00:00:00Z Latent-GRPO：面向潜在推理的稳定 RL 优化——GSM8K-Aug 上提升 7.86 分、AIME 上提升 4.27 分，推理链长度缩短 3-4 倍 https://24-ai.news/zh/news/2026-05-02/arxiv-mcphunt-mcp-credential-propagation/ 24 AI zh 2026-05-02T00:00:00Z MCPHunt：首个衡量多服务器MCP代理信任边界凭证泄露的基准测试——泄露率11.5%至41.3% https://24-ai.news/zh/news/2026-05-02/arxiv-msr-synthetic-computers-1000-scale/ 24 AI zh 2026-05-02T00:00:00Z Microsoft Research Synthetic Computers：1000台合成计算机作为长视程生产力AI代理训练的基础底层 https://24-ai.news/zh/news/2026-05-02/aws-transform-bi-migration-bedrock-agentcore/ 24 AI zh 2026-05-02T00:00:00Z AWS Transform利用AI代理将Tableau和Power BI仪表盘自动迁移至QuickSight，耗时从月缩短至天 https://24-ai.news/zh/news/2026-05-02/github-copilot-gpt-52-deprecation-june-2026/ 24 AI zh 2026-05-02T00:00:00Z GitHub将于2026年6月1日从Copilot中弃用GPT-5.2和GPT-5.2-Codex——迁移至GPT-5.5和GPT-5.3-Codex https://24-ai.news/zh/news/2026-05-02/google-research-open-science-250k-reach/ 24 AI zh 2026-05-02T00:00:00Z Google Research开源工具触及25万研究人员：从基因组学到为3800万农民提供季风预报 https://24-ai.news/zh/news/2026-05-02/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI zh 2026-05-02T00:00:00Z NIST CAISI对DeepSeek V4 Pro的评估：在5个领域9个基准测试中落后美国前沿模型8个月 https://24-ai.news/ja/news/2026-05-02/arxiv-exploration-hacking-rl-resistance/ 24 AI ja 2026-05-02T00:00:00Z 探索ハッキング:LLMはRL訓練に抵抗し、自らの能力を戦略的に抑制することを学べるのか https://24-ai.news/ja/news/2026-05-02/arxiv-kellybench-premier-league-decisions/ 24 AI ja 2026-05-02T00:00:00Z KellyBench：AIエージェントがPremier Leagueシーズンを通じて賭けの資金を管理——主要モデルはすべて損失 https://24-ai.news/ja/news/2026-05-02/arxiv-latent-grpo-reasoning-rl/ 24 AI ja 2026-05-02T00:00:00Z Latent-GRPO:潜在推論のための安定したRL最適化——GSM8K-Augで7.86ポイント、AIMEで4.27ポイント向上、推論チェーンは3-4倍短縮 https://24-ai.news/ja/news/2026-05-02/arxiv-mcphunt-mcp-credential-propagation/ 24 AI ja 2026-05-02T00:00:00Z MCPHunt：マルチサーバーMCPエージェントの信頼境界を越えた認証情報漏洩を測定する初のベンチマーク——漏洩率11.5〜41.3% https://24-ai.news/ja/news/2026-05-02/arxiv-msr-synthetic-computers-1000-scale/ 24 AI ja 2026-05-02T00:00:00Z Microsoft Research Synthetic Computers：長期生産性AIエージェントトレーニングの基盤として1,000台の合成コンピューターを活用 https://24-ai.news/ja/news/2026-05-02/aws-transform-bi-migration-bedrock-agentcore/ 24 AI ja 2026-05-02T00:00:00Z AWS TransformがTableauとPower BIのダッシュボードをQuickSightへ自動移行——数か月かかる作業を数日に短縮 https://24-ai.news/ja/news/2026-05-02/github-copilot-gpt-52-deprecation-june-2026/ 24 AI ja 2026-05-02T00:00:00Z GitHubが2026年6月1日にCopilotからGPT-5.2とGPT-5.2-Codexを廃止——GPT-5.5とGPT-5.3-Codexへの移行 https://24-ai.news/ja/news/2026-05-02/google-research-open-science-250k-reach/ 24 AI ja 2026-05-02T00:00:00Z Google Researchのオープンソースツールが25万人の研究者に到達：ゲノミクスから3,800万人の農家への季節風予報まで https://24-ai.news/ja/news/2026-05-02/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI ja 2026-05-02T00:00:00Z NIST CAISI によるDeepSeek V4 Pro評価：5分野9ベンチマークで米国フロンティアモデルに8か月遅れ https://24-ai.news/ko/news/2026-05-02/arxiv-exploration-hacking-rl-resistance/ 24 AI ko 2026-05-02T00:00:00Z 탐색 해킹: LLM은 RL 훈련에 저항하고 자신의 능력을 전략적으로 억제하는 법을 배울 수 있는가? https://24-ai.news/ko/news/2026-05-02/arxiv-kellybench-premier-league-decisions/ 24 AI ko 2026-05-02T00:00:00Z KellyBench: AI 에이전트가 Premier League 시즌 내내 베팅 자금을 관리——모든 주요 모델이 손실 https://24-ai.news/ko/news/2026-05-02/arxiv-latent-grpo-reasoning-rl/ 24 AI ko 2026-05-02T00:00:00Z Latent-GRPO: 잠재 추론을 위한 안정적 RL 최적화 — GSM8K-Aug에서 7.86점, AIME에서 4.27점 향상, 추론 체인은 3-4배 단축 https://24-ai.news/ko/news/2026-05-02/arxiv-mcphunt-mcp-credential-propagation/ 24 AI ko 2026-05-02T00:00:00Z MCPHunt: 다중 서버 MCP 에이전트의 신뢰 경계 간 자격증명 유출을 측정하는 최초의 벤치마크——유출률 11.5~41.3% https://24-ai.news/ko/news/2026-05-02/arxiv-msr-synthetic-computers-1000-scale/ 24 AI ko 2026-05-02T00:00:00Z Microsoft Research Synthetic Computers: 장기 생산성 AI 에이전트 훈련의 기반으로서 1,000개의 합성 컴퓨터 https://24-ai.news/ko/news/2026-05-02/aws-transform-bi-migration-bedrock-agentcore/ 24 AI ko 2026-05-02T00:00:00Z AWS Transform, Tableau 및 Power BI 대시보드를 QuickSight로 자동 마이그레이션——수개월 작업을 수일로 단축 https://24-ai.news/ko/news/2026-05-02/github-copilot-gpt-52-deprecation-june-2026/ 24 AI ko 2026-05-02T00:00:00Z GitHub, 2026년 6월 1일 Copilot에서 GPT-5.2 및 GPT-5.2-Codex 지원 종료——GPT-5.5 및 GPT-5.3-Codex로 마이그레이션 https://24-ai.news/ko/news/2026-05-02/google-research-open-science-250k-reach/ 24 AI ko 2026-05-02T00:00:00Z Google Research 오픈소스 도구, 25만 명의 연구자에게 도달: 유전체학부터 3,800만 농부를 위한 몬순 예보까지 https://24-ai.news/ko/news/2026-05-02/nist-caisi-deepseek-v4-pro-evaluation/ 24 AI ko 2026-05-02T00:00:00Z NIST CAISI의 DeepSeek V4 Pro 평가: 5개 영역 9개 벤치마크에서 미국 프론티어 모델보다 8개월 뒤처짐