https://24-ai.news/hr/news/2026-06-30/allen-discoformer-density-score/ 24 AI hr 2026-06-30T00:00:00Z Allen Institute: DiScoFormer — jedan transformer za gustoću i score kroz različite distribucije https://24-ai.news/hr/news/2026-06-30/amd-rocm-low-latency-gemm-inference/ 24 AI hr 2026-06-30T00:00:00Z AMD: ROCm low-latency GEMM kerneli ubrzavaju LLM inferenciju do 1,79× na Instinct MI355X https://24-ai.news/hr/news/2026-06-30/arxiv-agent-native-immune-system/ 24 AI hr 2026-06-30T00:00:00Z arXiv:2606.28270: Agent-Native Immune System — šestoslojna runtime obrana ugrađena u zaključivanje AI agenata https://24-ai.news/hr/news/2026-06-30/arxiv-tandem-rl-verifiable-rewards/ 24 AI hr 2026-06-30T00:00:00Z arXiv:2606.28166: Tandem RL — verifikabilne nagrade uz čitljiviji lanac misli i bolji handoff na manji model https://24-ai.news/hr/news/2026-06-30/arxiv-toolprivacybench-agent-privacy/ 24 AI hr 2026-06-30T00:00:00Z arXiv:2606.28061: ToolPrivacyBench — mjeri 'need-to-know' privatnost u LLM agentima s alatima https://24-ai.news/hr/news/2026-06-30/arxiv-world-model-planning-agents/ 24 AI hr 2026-06-30T00:00:00Z arXiv:2606.27483: Internalizing the Future — jedinstvena paradigma treniranja za svjetski model planiranja LLM agenata https://24-ai.news/hr/news/2026-06-30/aws-multitenant-agent-row-level-security/ 24 AI hr 2026-06-30T00:00:00Z AWS: multi-tenant AI agent s row-level security i Split-Plane SQL kriptografskim granicama podataka https://24-ai.news/hr/news/2026-06-30/aws-nova-2-lite-claude-document-pipeline/ 24 AI hr 2026-06-30T00:00:00Z AWS: Nova 2 Lite i Claude — cost-optimizirani pipeline za obradu skeniranih dokumenata https://24-ai.news/hr/news/2026-06-30/github-copilot-claude-opus-4-8-fast/ 24 AI hr 2026-06-30T00:00:00Z GitHub: Claude Opus 4.8 fast mode stiže u Copilot preview; Anthropic ukida fast za Opus 4.6 https://24-ai.news/hr/news/2026-06-30/langchain-dynamic-subagents-deep-agents/ 24 AI hr 2026-06-30T00:00:00Z LangChain: Dynamic Subagents u Deep Agents — agent piše kod koji paralelno dispatcha stotine podagenata https://24-ai.news/hr/news/2026-06-30/meta-brain2qwerty-noninvasive-text/ 24 AI hr 2026-06-30T00:00:00Z Meta: Brain2Qwerty v2 — neinvazivno dekodiranje misli u tekst s 61% točnosti, bez kirurškog implantata https://24-ai.news/hr/news/2026-06-30/microsoft-agent-confidence-index-2026/ 24 AI hr 2026-06-30T00:00:00Z Microsoft: 2026 Agent Confidence Index — 300 graditelja, prosječno povjerenje u AI agente 64/100 https://24-ai.news/hr/news/2026-06-30/microsoft-memora-agent-memory-tokens/ 24 AI hr 2026-06-30T00:00:00Z Microsoft Research: Memora — memorija AI agenata uz do 98% manje tokena i SOTA na dugim razgovorima https://24-ai.news/hr/news/2026-06-30/nvidia-palantir-nemotron-sovereign-ai/ 24 AI hr 2026-06-30T00:00:00Z NVIDIA: Palantir i NVIDIA Nemotron donose suvereni AI američkim agencijama u air-gapped sustavima https://24-ai.news/hr/news/2026-06-30/openai-europe-ai-workforce-mapping/ 24 AI hr 2026-06-30T00:00:00Z OpenAI: Mapping Europe's AI Workforce — izvještaj o prilikama AI-ja za EU radnu snagu https://24-ai.news/en/news/2026-06-30/allen-discoformer-density-score/ 24 AI en 2026-06-30T00:00:00Z Allen Institute: DiScoFormer — One Transformer for Density and Score Across Distributions https://24-ai.news/en/news/2026-06-30/amd-rocm-low-latency-gemm-inference/ 24 AI en 2026-06-30T00:00:00Z AMD: ROCm Low-Latency GEMM Kernels Speed Up LLM Inference by Up to 1.79× on Instinct MI355X https://24-ai.news/en/news/2026-06-30/arxiv-agent-native-immune-system/ 24 AI en 2026-06-30T00:00:00Z arXiv:2606.28270: Agent-Native Immune System — Six-Layer Runtime Defense Built Into AI Agent Reasoning https://24-ai.news/en/news/2026-06-30/arxiv-tandem-rl-verifiable-rewards/ 24 AI en 2026-06-30T00:00:00Z arXiv:2606.28166: Tandem RL — Verifiable Rewards With More Readable Chain of Thought and Better Handoff to Smaller Models https://24-ai.news/en/news/2026-06-30/arxiv-toolprivacybench-agent-privacy/ 24 AI en 2026-06-30T00:00:00Z arXiv:2606.28061: ToolPrivacyBench — Measuring 'Need-to-Know' Privacy in LLM Agents With Tools https://24-ai.news/en/news/2026-06-30/arxiv-world-model-planning-agents/ 24 AI en 2026-06-30T00:00:00Z arXiv:2606.27483: Internalizing the Future — A Unified Training Paradigm for World Model Planning in LLM Agents https://24-ai.news/en/news/2026-06-30/aws-multitenant-agent-row-level-security/ 24 AI en 2026-06-30T00:00:00Z AWS: Multi-Tenant AI Agent With Row-Level Security and Split-Plane SQL Cryptographic Data Boundaries https://24-ai.news/en/news/2026-06-30/aws-nova-2-lite-claude-document-pipeline/ 24 AI en 2026-06-30T00:00:00Z AWS: Nova 2 Lite and Claude — Cost-Optimized Pipeline for Processing Scanned Documents https://24-ai.news/en/news/2026-06-30/github-copilot-claude-opus-4-8-fast/ 24 AI en 2026-06-30T00:00:00Z GitHub: Claude Opus 4.8 Fast Mode Arrives in Copilot Preview; Anthropic Retires Fast Mode for Opus 4.6 https://24-ai.news/en/news/2026-06-30/langchain-dynamic-subagents-deep-agents/ 24 AI en 2026-06-30T00:00:00Z LangChain: Dynamic Subagents in Deep Agents — Agent Writes Code That Dispatches Hundreds of Subagents in Parallel https://24-ai.news/en/news/2026-06-30/meta-brain2qwerty-noninvasive-text/ 24 AI en 2026-06-30T00:00:00Z Meta: Brain2Qwerty v2 — Non-Invasive Thought-to-Text Decoding at 61% Accuracy, Without Surgical Implants https://24-ai.news/en/news/2026-06-30/microsoft-agent-confidence-index-2026/ 24 AI en 2026-06-30T00:00:00Z Microsoft: 2026 Agent Confidence Index — 300 Builders, Average Confidence in AI Agents 64/100 https://24-ai.news/en/news/2026-06-30/microsoft-memora-agent-memory-tokens/ 24 AI en 2026-06-30T00:00:00Z Microsoft Research: Memora — AI Agent Memory With Up to 98% Fewer Tokens and SOTA on Long Conversations https://24-ai.news/en/news/2026-06-30/nvidia-palantir-nemotron-sovereign-ai/ 24 AI en 2026-06-30T00:00:00Z NVIDIA: Palantir and NVIDIA Nemotron Bring Sovereign AI to US Agencies in Air-Gapped Systems https://24-ai.news/en/news/2026-06-30/openai-europe-ai-workforce-mapping/ 24 AI en 2026-06-30T00:00:00Z OpenAI: Mapping Europe's AI Workforce — Report on AI Opportunities for the EU Workforce https://24-ai.news/de/news/2026-06-30/allen-discoformer-density-score/ 24 AI de 2026-06-30T00:00:00Z Allen Institute: DiScoFormer — ein Transformer für Dichte und Score über verschiedene Verteilungen https://24-ai.news/de/news/2026-06-30/amd-rocm-low-latency-gemm-inference/ 24 AI de 2026-06-30T00:00:00Z AMD: ROCm Low-Latency-GEMM-Kernel beschleunigen LLM-Inferenz bis zu 1,79× auf Instinct MI355X https://24-ai.news/de/news/2026-06-30/arxiv-agent-native-immune-system/ 24 AI de 2026-06-30T00:00:00Z arXiv:2606.28270: Agent-Native Immune System — sechsschichtige Runtime-Abwehr in der KI-Agenten-Kognitionsschleife https://24-ai.news/de/news/2026-06-30/arxiv-tandem-rl-verifiable-rewards/ 24 AI de 2026-06-30T00:00:00Z arXiv:2606.28166: Tandem RL — verifizierbare Belohnungen mit besser lesbarer Gedankenkette und verbessertem Handoff auf kleinere Modelle https://24-ai.news/de/news/2026-06-30/arxiv-toolprivacybench-agent-privacy/ 24 AI de 2026-06-30T00:00:00Z arXiv:2606.28061: ToolPrivacyBench — misst „Need-to-Know”-Datenschutz in LLM-Agenten mit Werkzeugen https://24-ai.news/de/news/2026-06-30/arxiv-world-model-planning-agents/ 24 AI de 2026-06-30T00:00:00Z arXiv:2606.27483: Internalizing the Future — einheitliches Training-Paradigma für World-Model-Planning von LLM-Agenten https://24-ai.news/de/news/2026-06-30/aws-multitenant-agent-row-level-security/ 24 AI de 2026-06-30T00:00:00Z AWS: Multi-Tenant-KI-Agent mit Row-Level-Security und Split-Plane-SQL als kryptographische Datengrenzen https://24-ai.news/de/news/2026-06-30/aws-nova-2-lite-claude-document-pipeline/ 24 AI de 2026-06-30T00:00:00Z AWS: Nova 2 Lite und Claude — kostenoptimierte Pipeline zur Verarbeitung gescannter Dokumente https://24-ai.news/de/news/2026-06-30/github-copilot-claude-opus-4-8-fast/ 24 AI de 2026-06-30T00:00:00Z GitHub: Claude Opus 4.8 Fast Mode kommt in Copilot-Preview; Anthropic stellt Fast für Opus 4.6 ein https://24-ai.news/de/news/2026-06-30/langchain-dynamic-subagents-deep-agents/ 24 AI de 2026-06-30T00:00:00Z LangChain: Dynamic Subagents in Deep Agents — Agent schreibt Code zum parallelen Dispatch hunderter Unteragenten https://24-ai.news/de/news/2026-06-30/meta-brain2qwerty-noninvasive-text/ 24 AI de 2026-06-30T00:00:00Z Meta: Brain2Qwerty v2 — nichtinvasive Dekodierung von Gedanken in Text mit 61% Genauigkeit, ohne chirurgisches Implantat https://24-ai.news/de/news/2026-06-30/microsoft-agent-confidence-index-2026/ 24 AI de 2026-06-30T00:00:00Z Microsoft: 2026 Agent Confidence Index — 300 Entwickler, durchschnittliches Vertrauen in KI-Agenten 64/100 https://24-ai.news/de/news/2026-06-30/microsoft-memora-agent-memory-tokens/ 24 AI de 2026-06-30T00:00:00Z Microsoft Research: Memora — KI-Agenten-Gedächtnis mit bis zu 98% weniger Token und SOTA bei langen Gesprächen https://24-ai.news/de/news/2026-06-30/nvidia-palantir-nemotron-sovereign-ai/ 24 AI de 2026-06-30T00:00:00Z NVIDIA: Palantir und NVIDIA Nemotron bringen souveräne KI für US-Behörden in Air-Gapped-Systemen https://24-ai.news/de/news/2026-06-30/openai-europe-ai-workforce-mapping/ 24 AI de 2026-06-30T00:00:00Z OpenAI: Mapping Europe's AI Workforce — Bericht über KI-Chancen für EU-Arbeitskräfte https://24-ai.news/zh/news/2026-06-30/allen-discoformer-density-score/ 24 AI zh 2026-06-30T00:00:00Z Allen Institute：DiScoFormer——单个Transformer跨分布同步估计密度与得分函数 https://24-ai.news/zh/news/2026-06-30/amd-rocm-low-latency-gemm-inference/ 24 AI zh 2026-06-30T00:00:00Z AMD：ROCm低延迟GEMM内核在Instinct MI355X上将LLM推理提速最高1.79倍 https://24-ai.news/zh/news/2026-06-30/arxiv-agent-native-immune-system/ 24 AI zh 2026-06-30T00:00:00Z arXiv:2606.28270：Agent原生免疫系统——嵌入AI智能体推理中的六层运行时防御 https://24-ai.news/zh/news/2026-06-30/arxiv-tandem-rl-verifiable-rewards/ 24 AI zh 2026-06-30T00:00:00Z arXiv:2606.28166：Tandem RL——可验证奖励提升思维链可读性并改善向小模型的迁移 https://24-ai.news/zh/news/2026-06-30/arxiv-toolprivacybench-agent-privacy/ 24 AI zh 2026-06-30T00:00:00Z arXiv:2606.28061：ToolPrivacyBench——衡量LLM工具调用智能体的「知其所需」隐私保护 https://24-ai.news/zh/news/2026-06-30/arxiv-world-model-planning-agents/ 24 AI zh 2026-06-30T00:00:00Z arXiv:2606.27483：内化未来——LLM智能体世界模型规划的统一训练范式 https://24-ai.news/zh/news/2026-06-30/aws-multitenant-agent-row-level-security/ 24 AI zh 2026-06-30T00:00:00Z AWS：多租户AI智能体结合行级安全与Split-Plane SQL实现加密数据隔离 https://24-ai.news/zh/news/2026-06-30/aws-nova-2-lite-claude-document-pipeline/ 24 AI zh 2026-06-30T00:00:00Z AWS：Nova 2 Lite与Claude——扫描文档批量处理的成本优化双阶段流水线 https://24-ai.news/zh/news/2026-06-30/github-copilot-claude-opus-4-8-fast/ 24 AI zh 2026-06-30T00:00:00Z GitHub：Claude Opus 4.8快速模式进入Copilot预览；Anthropic终止Opus 4.6的快速模式 https://24-ai.news/zh/news/2026-06-30/langchain-dynamic-subagents-deep-agents/ 24 AI zh 2026-06-30T00:00:00Z LangChain：Deep Agents中的Dynamic Subagents——智能体编写代码并行调度数百个子智能体 https://24-ai.news/zh/news/2026-06-30/meta-brain2qwerty-noninvasive-text/ 24 AI zh 2026-06-30T00:00:00Z Meta：Brain2Qwerty v2——无需外科植入物，以61%准确率将大脑信号转化为文字 https://24-ai.news/zh/news/2026-06-30/microsoft-agent-confidence-index-2026/ 24 AI zh 2026-06-30T00:00:00Z Microsoft：2026智能体信心指数——300位构建者对AI智能体的平均信心得分64/100 https://24-ai.news/zh/news/2026-06-30/microsoft-memora-agent-memory-tokens/ 24 AI zh 2026-06-30T00:00:00Z Microsoft Research：Memora——减少最多98%Token消耗并在长对话上达到SOTA的AI智能体记忆框架 https://24-ai.news/zh/news/2026-06-30/nvidia-palantir-nemotron-sovereign-ai/ 24 AI zh 2026-06-30T00:00:00Z NVIDIA：Palantir与NVIDIA Nemotron为气隙隔离系统中的美国政府机构带来主权AI https://24-ai.news/zh/news/2026-06-30/openai-europe-ai-workforce-mapping/ 24 AI zh 2026-06-30T00:00:00Z OpenAI：欧洲AI劳动力地图——关于AI对欧盟劳动力机遇影响的报告 https://24-ai.news/ja/news/2026-06-30/allen-discoformer-density-score/ 24 AI ja 2026-06-30T00:00:00Z Allen Institute: DiScoFormer — 一つのTransformerが様々な分布の密度とスコアを同時推定 https://24-ai.news/ja/news/2026-06-30/amd-rocm-low-latency-gemm-inference/ 24 AI ja 2026-06-30T00:00:00Z AMD: ROCm低レイテンシGEMMカーネルがInstinct MI355XのLLM推論を最大1.79倍高速化 https://24-ai.news/ja/news/2026-06-30/arxiv-agent-native-immune-system/ 24 AI ja 2026-06-30T00:00:00Z arXiv:2606.28270: Agent-Native Immune System — AIエージェントの推論に組み込まれた6層ランタイム防御 https://24-ai.news/ja/news/2026-06-30/arxiv-tandem-rl-verifiable-rewards/ 24 AI ja 2026-06-30T00:00:00Z arXiv:2606.28166: Tandem RL — 検証可能な報酬により思考の連鎖の可読性と小モデルへのハンドオフを改善 https://24-ai.news/ja/news/2026-06-30/arxiv-toolprivacybench-agent-privacy/ 24 AI ja 2026-06-30T00:00:00Z arXiv:2606.28061: ToolPrivacyBench — ツールを使うLLMエージェントの「need-to-know」プライバシーを計測 https://24-ai.news/ja/news/2026-06-30/arxiv-world-model-planning-agents/ 24 AI ja 2026-06-30T00:00:00Z arXiv:2606.27483: Internalizing the Future — LLMエージェントの世界モデル計画のための統一的なトレーニングパラダイム https://24-ai.news/ja/news/2026-06-30/aws-multitenant-agent-row-level-security/ 24 AI ja 2026-06-30T00:00:00Z AWS: 行レベルセキュリティとSplit-Plane SQLの暗号化データ境界を持つマルチテナントAIエージェント https://24-ai.news/ja/news/2026-06-30/aws-nova-2-lite-claude-document-pipeline/ 24 AI ja 2026-06-30T00:00:00Z AWS: Nova 2 Liteとクロード — スキャン文書処理のコスト最適化パイプライン https://24-ai.news/ja/news/2026-06-30/github-copilot-claude-opus-4-8-fast/ 24 AI ja 2026-06-30T00:00:00Z GitHub: Claude Opus 4.8 fastモードがCopilotのプレビューに登場、AnthropicはOpus 4.6のfastモードを廃止 https://24-ai.news/ja/news/2026-06-30/langchain-dynamic-subagents-deep-agents/ 24 AI ja 2026-06-30T00:00:00Z LangChain: Deep AgentsのDynamic Subagents — エージェントが数百のサブエージェントを並列ディスパッチするコードを記述 https://24-ai.news/ja/news/2026-06-30/meta-brain2qwerty-noninvasive-text/ 24 AI ja 2026-06-30T00:00:00Z Meta: Brain2Qwerty v2 — 外科的インプラントなしで脳信号を61%の精度でテキストにデコード https://24-ai.news/ja/news/2026-06-30/microsoft-agent-confidence-index-2026/ 24 AI ja 2026-06-30T00:00:00Z Microsoft: 2026 Agent Confidence Index — 300人のビルダーがAIエージェントへの平均信頼度64/100と評価 https://24-ai.news/ja/news/2026-06-30/microsoft-memora-agent-memory-tokens/ 24 AI ja 2026-06-30T00:00:00Z Microsoft Research: Memora — AIエージェントの記憶を最大98%少ないトークンとSOTAの長期会話で実現 https://24-ai.news/ja/news/2026-06-30/nvidia-palantir-nemotron-sovereign-ai/ 24 AI ja 2026-06-30T00:00:00Z NVIDIA: PalantirとNVIDIA Nemotronがエアギャップシステムの米国政府機関に主権AIを提供 https://24-ai.news/ja/news/2026-06-30/openai-europe-ai-workforce-mapping/ 24 AI ja 2026-06-30T00:00:00Z OpenAI: EUの労働力へのAIの機会をまとめた「Mapping Europe's AI Workforce」レポート https://24-ai.news/ko/news/2026-06-30/allen-discoformer-density-score/ 24 AI ko 2026-06-30T00:00:00Z Allen Institute: DiScoFormer — 하나의 트랜스포머로 다양한 분포의 밀도와 점수 추정 https://24-ai.news/ko/news/2026-06-30/amd-rocm-low-latency-gemm-inference/ 24 AI ko 2026-06-30T00:00:00Z AMD: ROCm 저지연 GEMM 커널로 Instinct MI355X에서 LLM 추론 최대 1.79배 가속 https://24-ai.news/ko/news/2026-06-30/arxiv-agent-native-immune-system/ 24 AI ko 2026-06-30T00:00:00Z arXiv:2606.28270: Agent-Native Immune System — AI 에이전트 추론에 내장된 6계층 런타임 방어 https://24-ai.news/ko/news/2026-06-30/arxiv-tandem-rl-verifiable-rewards/ 24 AI ko 2026-06-30T00:00:00Z arXiv:2606.28166: Tandem RL — 더 읽기 쉬운 사고 연쇄와 소형 모델 핸드오프를 갖춘 검증 가능한 보상 https://24-ai.news/ko/news/2026-06-30/arxiv-toolprivacybench-agent-privacy/ 24 AI ko 2026-06-30T00:00:00Z arXiv:2606.28061: ToolPrivacyBench — 도구를 사용하는 LLM 에이전트의 『필요한 만큼만』 프라이버시 측정 https://24-ai.news/ko/news/2026-06-30/arxiv-world-model-planning-agents/ 24 AI ko 2026-06-30T00:00:00Z arXiv:2606.27483: Internalizing the Future — LLM 에이전트 세계 모델 계획을 위한 통합 에이전트 학습 패러다임 https://24-ai.news/ko/news/2026-06-30/aws-multitenant-agent-row-level-security/ 24 AI ko 2026-06-30T00:00:00Z AWS: 행 수준 보안과 Split-Plane SQL 암호화 데이터 경계를 갖춘 다중 테넌트 AI 에이전트 https://24-ai.news/ko/news/2026-06-30/aws-nova-2-lite-claude-document-pipeline/ 24 AI ko 2026-06-30T00:00:00Z AWS: Nova 2 Lite와 Claude — 스캔 문서 처리를 위한 비용 최적화 파이프라인 https://24-ai.news/ko/news/2026-06-30/github-copilot-claude-opus-4-8-fast/ 24 AI ko 2026-06-30T00:00:00Z GitHub: Claude Opus 4.8 fast mode가 Copilot 프리뷰에 출시; Anthropic은 Opus 4.6의 fast 모드 종료 https://24-ai.news/ko/news/2026-06-30/langchain-dynamic-subagents-deep-agents/ 24 AI ko 2026-06-30T00:00:00Z LangChain: Deep Agents의 Dynamic Subagents — 에이전트가 수백 개의 하위 에이전트를 병렬로 디스패치하는 코드 작성 https://24-ai.news/ko/news/2026-06-30/meta-brain2qwerty-noninvasive-text/ 24 AI ko 2026-06-30T00:00:00Z Meta: Brain2Qwerty v2 — 수술적 임플란트 없이 61% 정확도로 생각을 텍스트로 비침습적 해독 https://24-ai.news/ko/news/2026-06-30/microsoft-agent-confidence-index-2026/ 24 AI ko 2026-06-30T00:00:00Z Microsoft: 2026 Agent Confidence Index — 300명의 개발자, AI 에이전트에 대한 평균 신뢰도 64/100 https://24-ai.news/ko/news/2026-06-30/microsoft-memora-agent-memory-tokens/ 24 AI ko 2026-06-30T00:00:00Z Microsoft Research: Memora — 토큰 최대 98% 감소와 장문 대화 SOTA를 달성하는 AI 에이전트 메모리 https://24-ai.news/ko/news/2026-06-30/nvidia-palantir-nemotron-sovereign-ai/ 24 AI ko 2026-06-30T00:00:00Z NVIDIA: Palantir과 NVIDIA Nemotron이 에어갭 시스템의 미국 기관에 주권 AI 제공 https://24-ai.news/ko/news/2026-06-30/openai-europe-ai-workforce-mapping/ 24 AI ko 2026-06-30T00:00:00Z OpenAI: Mapping Europe's AI Workforce — EU 노동력을 위한 AI 기회에 관한 보고서 https://24-ai.news/hr/news/2026-06-28/amd-rocm-gpu-workload-preemption/ 24 AI hr 2026-06-28T00:00:00Z AMD: Resource Manager automatski istiskuje neaktivne GPU workloadove i vraća resurse u zajednički pool klastera https://24-ai.news/hr/news/2026-06-28/arxiv-cot-training-action-prediction/ 24 AI hr 2026-06-28T00:00:00Z arXiv:2606.26935: dobici CoT treninga slijevaju se u jače predviđanje akcije, a ne u dublje rezoniranje agenata https://24-ai.news/hr/news/2026-06-28/arxiv-evaf-parametric-agent-memory/ 24 AI hr 2026-06-28T00:00:00Z arXiv:2606.26806: EVAF konsolidira pamćenje agenata u parametre modela za postojanost cilja nakon brisanja konteksta https://24-ai.news/hr/news/2026-06-28/arxiv-reasoning-tokens-failure-divergence/ 24 AI hr 2026-06-28T00:00:00Z arXiv:2606.26502: reasoning modeli troše više tokena na zadatke koje pogriješe, suprotno od ljudi koji odustaju https://24-ai.news/hr/news/2026-06-28/github-copilot-agentic-harness-benchmark/ 24 AI hr 2026-06-28T00:00:00Z GitHub: Copilot agentic harness dostiže razinu vendor-harnessa uz manju potrošnju tokena kroz 20+ frontier modela https://24-ai.news/hr/news/2026-06-28/github-mai-code-1-flash-copilot-ga/ 24 AI hr 2026-06-28T00:00:00Z GitHub: MAI-Code-1-Flash, Microsoftov coding model, sada općedostupan u Copilot Business i Enterprise planovima https://24-ai.news/en/news/2026-06-28/amd-rocm-gpu-workload-preemption/ 24 AI en 2026-06-28T00:00:00Z AMD: Resource Manager automatically preempts inactive GPU workloads and returns resources to the cluster shared pool https://24-ai.news/en/news/2026-06-28/arxiv-cot-training-action-prediction/ 24 AI en 2026-06-28T00:00:00Z arXiv:2606.26935: CoT training gains land in stronger action prediction, not deeper agent reasoning https://24-ai.news/en/news/2026-06-28/arxiv-evaf-parametric-agent-memory/ 24 AI en 2026-06-28T00:00:00Z arXiv:2606.26806: EVAF consolidates agent memory into model parameters for goal persistence after context deletion https://24-ai.news/en/news/2026-06-28/arxiv-reasoning-tokens-failure-divergence/ 24 AI en 2026-06-28T00:00:00Z arXiv:2606.26502: reasoning models spend more tokens on tasks they fail, opposite to humans who disengage https://24-ai.news/en/news/2026-06-28/github-copilot-agentic-harness-benchmark/ 24 AI en 2026-06-28T00:00:00Z GitHub: Copilot agentic harness reaches vendor-harness level with lower token consumption across 20+ frontier models https://24-ai.news/en/news/2026-06-28/github-mai-code-1-flash-copilot-ga/ 24 AI en 2026-06-28T00:00:00Z GitHub: MAI-Code-1-Flash, Microsoft's coding model, now generally available in Copilot Business and Enterprise plans https://24-ai.news/de/news/2026-06-28/amd-rocm-gpu-workload-preemption/ 24 AI de 2026-06-28T00:00:00Z AMD: Resource Manager verdrängt automatisch inaktive GPU-Workloads und gibt Ressourcen an den Cluster-Pool zurück https://24-ai.news/de/news/2026-06-28/arxiv-cot-training-action-prediction/ 24 AI de 2026-06-28T00:00:00Z arXiv:2606.26935: CoT-Trainingsgewinne fließen in stärkere Aktionsvorhersage, nicht in tieferes Agenten-Reasoning https://24-ai.news/de/news/2026-06-28/arxiv-evaf-parametric-agent-memory/ 24 AI de 2026-06-28T00:00:00Z arXiv:2606.26806: EVAF konsolidiert Agentenspeicher in Modellparameter für Zielbeständigkeit nach Kontextlöschung https://24-ai.news/de/news/2026-06-28/arxiv-reasoning-tokens-failure-divergence/ 24 AI de 2026-06-28T00:00:00Z arXiv:2606.26502: Reasoning-Modelle verbrauchen mehr Tokens bei Fehlern – im Gegensatz zu Menschen, die aufgeben https://24-ai.news/de/news/2026-06-28/github-copilot-agentic-harness-benchmark/ 24 AI de 2026-06-28T00:00:00Z GitHub: Copilot Agentic Harness erreicht Vendor-Niveau mit geringerem Token-Verbrauch über 20+ Frontier-Modelle https://24-ai.news/de/news/2026-06-28/github-mai-code-1-flash-copilot-ga/ 24 AI de 2026-06-28T00:00:00Z GitHub: MAI-Code-1-Flash, Microsofts Coding-Modell, jetzt allgemein verfügbar in Copilot Business und Enterprise https://24-ai.news/zh/news/2026-06-28/amd-rocm-gpu-workload-preemption/ 24 AI zh 2026-06-28T00:00:00Z AMD: Resource Manager 自动抢占空闲 GPU 工作负载，将资源归还集群共享池 https://24-ai.news/zh/news/2026-06-28/arxiv-cot-training-action-prediction/ 24 AI zh 2026-06-28T00:00:00Z arXiv:2606.26935: CoT 训练收益流向更强的动作预测，而非更深的智能体推理 https://24-ai.news/zh/news/2026-06-28/arxiv-evaf-parametric-agent-memory/ 24 AI zh 2026-06-28T00:00:00Z arXiv:2606.26806: EVAF 将智能体记忆固化入模型参数，在上下文清除后保持目标持久性 https://24-ai.news/zh/news/2026-06-28/arxiv-reasoning-tokens-failure-divergence/ 24 AI zh 2026-06-28T00:00:00Z arXiv:2606.26502: 推理模型在出错任务上消耗更多 token，与人类放弃的行为相反 https://24-ai.news/zh/news/2026-06-28/github-copilot-agentic-harness-benchmark/ 24 AI zh 2026-06-28T00:00:00Z GitHub: Copilot 智能体 harness 在 20 余款前沿模型上达到供应商 harness 水平，且 token 消耗更低 https://24-ai.news/zh/news/2026-06-28/github-mai-code-1-flash-copilot-ga/ 24 AI zh 2026-06-28T00:00:00Z GitHub: MAI-Code-1-Flash，微软编程模型，现已在 Copilot Business 和 Enterprise 方案中正式发布 https://24-ai.news/ja/news/2026-06-28/amd-rocm-gpu-workload-preemption/ 24 AI ja 2026-06-28T00:00:00Z AMD: Resource Managerが非アクティブなGPUワークロードを自動的にプリエンプションしクラスタの共有プールにリソースを返還 https://24-ai.news/ja/news/2026-06-28/arxiv-cot-training-action-prediction/ 24 AI ja 2026-06-28T00:00:00Z arXiv:2606.26935：CoTトレーニングの恩恵はより深い推論ではなくより強い行動予測として現れる https://24-ai.news/ja/news/2026-06-28/arxiv-evaf-parametric-agent-memory/ 24 AI ja 2026-06-28T00:00:00Z arXiv:2606.26806：EVAFがエージェントの記憶をモデルパラメータに統合しコンテキスト削除後も目標を維持 https://24-ai.news/ja/news/2026-06-28/arxiv-reasoning-tokens-failure-divergence/ 24 AI ja 2026-06-28T00:00:00Z arXiv:2606.26502：推論モデルは失敗したタスクにより多くのトークンを消費し、諦める人間とは逆の傾向を示す https://24-ai.news/ja/news/2026-06-28/github-copilot-agentic-harness-benchmark/ 24 AI ja 2026-06-28T00:00:00Z GitHub: Copilot agenticハーネスが20以上のフロンティアモデルにわたってベンダーハーネス水準の性能を達成し、より少ないトークン消費を実現 https://24-ai.news/ja/news/2026-06-28/github-mai-code-1-flash-copilot-ga/ 24 AI ja 2026-06-28T00:00:00Z GitHub: MAI-Code-1-Flash、MicrosoftのコーディングモデルがCopilot BusinessおよびEnterpriseプランで一般提供開始 https://24-ai.news/ko/news/2026-06-28/amd-rocm-gpu-workload-preemption/ 24 AI ko 2026-06-28T00:00:00Z AMD: Resource Manager, 비활성 GPU 워크로드를 자동 선점하여 클러스터 공유 풀에 자원 반환 https://24-ai.news/ko/news/2026-06-28/arxiv-cot-training-action-prediction/ 24 AI ko 2026-06-28T00:00:00Z arXiv:2606.26935: CoT 훈련의 이득은 깊은 에이전트 추론이 아닌 행동 예측 강화로 귀결 https://24-ai.news/ko/news/2026-06-28/arxiv-evaf-parametric-agent-memory/ 24 AI ko 2026-06-28T00:00:00Z arXiv:2606.26806: EVAF, 컨텍스트 삭제 후에도 목표를 유지하는 에이전트 파라메트릭 메모리 통합 https://24-ai.news/ko/news/2026-06-28/arxiv-reasoning-tokens-failure-divergence/ 24 AI ko 2026-06-28T00:00:00Z arXiv:2606.26502: 추론 모델은 틀린 문제에 더 많은 토큰 소비 — 포기하는 사람과 반대 패턴 https://24-ai.news/ko/news/2026-06-28/github-copilot-agentic-harness-benchmark/ 24 AI ko 2026-06-28T00:00:00Z GitHub: Copilot 에이전트 하니스, 20개 이상의 프론티어 모델에서 벤더 하니스 수준 달성 및 토큰 사용량 절감 https://24-ai.news/ko/news/2026-06-28/github-mai-code-1-flash-copilot-ga/ 24 AI ko 2026-06-28T00:00:00Z GitHub: MAI-Code-1-Flash, 마이크로소프트의 코딩 모델, 이제 Copilot Business 및 Enterprise 플랜에서 일반 제공