Sierra 发布 Horizon 平台,让智能体能够追求长周期目标,例如发起贷款或为医疗程序获取事先授权,并可在数天、数周或数月内主动与客户互动以促成销售、升级订阅或预约试驾。Horizon 为 Agent OS 带来长周期规划与灵活的上下文引擎,每次互动都会让智能体更智能,且按业务成果而非 token 计费。
Hugging Face 披露,恶意数据集利用两条代码执行路径入侵处理 Worker,攻击者随后获取节点权限并横向进入多个内部集群。公司已修复漏洞、重建受影响节点并轮换凭据;其分析智能体使用 GLM-5.2 处理逾 17,000 条攻击事件,在数小时内完成通常需数天的取证工作。
析智能体使用 GLM-5.2 处理逾 17,000 条攻击事件,在数小时内完成通常需数天的取证工作。 · Hugging Face disclosed that a malicious dataset exploited two code-execution paths to compromise a processing worker, after which the attacker gained node-level access and moved laterally into several internal clusters. The company fixed the vulnerabilities, rebuilt affected nodes, and rotated credentials; its analysis agents used GLM-5.2 to examine more than 17,000 attack events, completing in hours forensic work that would usually take days.
和 Apache TVM FFI,并为智能体构建、基准测试和迭代优化 kernel 提供基础。 · Hugging Face introduced a dedicated Hub repository type for Kernels and strengthened security with trusted publishers, reproducible builds, and Sigstore cosign code signing; automatic signature verification during loading is not yet enabled. The update also revamps the kernels and kernel-builder CLIs, adds support for the Torch Stable ABI and Apache TVM FFI, and lays the foundation for agents to build, benchmark, and iteratively optimize kernels.
作者主张,范围明确、任务有限的 AI 应采用专业化模型或智能体,以更有效地利用参数,并可能减少模型幻觉与 Token 消耗。开放式任务更适合覆盖广泛但深度较浅的通用模型。专业知识收益何时触底目前难以衡量,作者希望借助机制可解释性进一步研究。
消耗。开放式任务更适合覆盖广泛但深度较浅的通用模型。专业知识收益何时触底目前难以衡量,作者希望借助机制可解释性进一步研究。 · The author argues that AI systems with a defined scope and a limited set of tasks should use specialized models or agents to employ parameters more efficiently and potentially reduce hallucinations and Token consumption. Open-ended tasks are better suited to general-purpose models with broad but shallower coverage. The point at which additional domain knowledge stops paying off remains difficult to measure, and the author hopes to
Sakana AI 与有限責任あずさ監査法人联合发布 CoffeeBench,这是一个模拟咖啡行业供应链的多智能体经济环境,让 LLM 智能体在 90 天内经营烘焙店并最大化净利润。实验中 GPT-5.5、Claude Opus 4.7 等模型均超过被动基线,但 Claude Haiku 4.5 出现亏损,并在三次试验中均出现「思考与行动脱节」的停滞现象。该研究将在 ICML2026 Workshop Failure Modes in Agentic AI 发表。