Meta Muse Wonder Full升至Agent Arena第七
Muse Wonder Full以【11.3%净提升】排名第七，Meta升至实验室第四；工具幻觉测试未发现问题。[1]
大黑AI速报
本期聚焦Meta Muse智能体排名跃升，以及AWSGoogle密集扩展智能体、模型部署与场景化工具。
模型动态
产品工具
AWS Agent Registry支持智能体运行时发现工具
Agent Registry支持【运行时工具发现】，减少API硬编码并强化企业工具治理。https://aws.amazon.com/blogs/machine-learning/discover-tools-at-runtime-with-aws-agent-registry/[4]
加拿大客户可通过Bedrock跨区调用OpenAI模型
加拿大客户现可借助跨区域推理访问OpenAI GPT模型，但仍需评估数据驻留合规。https://aws.amazon.com/blogs/machine-learning/access-openai-gpt-models-on-amazon-bedrock-from-canada-with-cross-region-inference/[5]
Gemini将个性化免手操作能力扩展至汽车
Gemini进入全球车载场景，结合个人上下文提供免手操作与个性化协助。https://blog.google/products-and-platforms/platforms/android/android-auto-cars-with-google-built-in-gemini-personal-intelligence/[6]
Google为小企业推出专属Gemini工具入口
Google推出面向小企业主的Gemini工具与商业入口，覆盖【营销】和日常运营。https://blog.google/innovation-and-ai/products/gemini-app/new-tools-small-business-owners/[8]
技巧教程
AWS展示线性注意力模型训练与部署方案
AWS用Trainium2训练部署Laplace 3B，摆脱随上下文增长的KV cache。https://aws.amazon.com/blogs/machine-learning/beyond-the-kv-cache-training-and-serving-i14s-laplace-3b-linear-attention-model-on-aws-trainium2-with-native-pytorch/[3]
AWS分享GPT-OSS-20B多语言推理微调方法
AWS演示在SageMaker HyperPod微调GPT-OSS-20B，覆盖四种欧洲语言推理。https://aws.amazon.com/blogs/machine-learning/fine-tuning-gpt-oss-20b-for-multilingual-reasoning-on-amazon-sagemaker-hyperpod/[7]
行业资讯
FairStress提出大模型决策公平性压力测试
FairStress评估模型在决策任务中的【公平性】与脆弱点，补足平均指标盲区。https://huggingface.co/blog/allenai/fairstress[2]