Briefing · 2026-10-08
技术简报 - 2026-10-08
On this page6 sections
The briefing's title, the editors' summary and their notes are written in Chinese, as are skill and knowledge entries.
Overview
#这一期的六条信号，都要分清两种数字：自己报的，和核验过的。ThinkingBox 按数据库最终状态给智能体评分，做错的任务里有三分之二没有报错就结束了。OpenAI 公开 719 篇数学结果，有 Lean 证明的只占四成，第二天就撤回了三篇。Nemotron 的两枚金牌，IMO 由官方评分，IOI 是自测。Claude Haiku 5.5 的标价降到十分之一，新分词器又多用两成五 token。另有两个开放模型：把文字、图片、音频放进同一向量空间的 EmbeddingGemma 2，以及与 Jev 互有胜负的决策模型 Clef。
Immediate attention
#The highest-rated published signals; the editors' pinned items come first.
The Agent Said It Was Done. The Database Disagreed.
The Agent Said It Was Done. The Database Disagreed.
For 让智能体直接改写业务数据的工程团队、负责智能体上线验收和监控的平台与运维团队、在多个模型之间选型、需要看可靠性而不只是平均分的决策者
Claude Haiku 5.5
As previously promised, here's Anthropic's new fast, low cost model: Introducing Claude Haiku 5.5. The previous Haiku, 4.5, was very much showing its age. It came out almost a year ago...
For 用小模型做分类、摘要、子任务分发的工程团队、负责模型路由、在 Haiku 与 Sonnet 之间分配流量的平台团队、处理长文档、提示常超过 10 万 token 的应用
Sharing AI progress in mathematics
OpenAI publishes new results on open problems in mathematics from an internal frontier model and shares Lean proof formalizations and research details on GitHub.
For 关注 AI 在科研中能做什么的研究者与技术决策者、做形式化验证、Lean 或定理证明相关工作的工程师、需要判断「AI 解决难题」类报道可信度的读者
Worth tracking
#Worth following, but each needs more context or confirmation before it becomes a priority.
One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO
One Model Family, Two Gold-Level Results: Fine-Tuning Nemotron for IOI and IMO
For 做模型后训练、数学或代码推理的研究与工程团队、想在自己任务上加验证循环的应用团队、需要判断竞赛类成绩可信度的读者
EmbeddingGemma 2: an open, lightweight multimodal embedding model
EmbeddingGemma 2: an open, lightweight multimodal embedding model
For 搭检索增强生成或语义搜索的工程团队、做代码检索、代码库问答的开发工具团队、需要在设备上做多模态检索的移动与端侧应用
Ollama v0.35.1: Clef decision models
Clef decision models Ollama now supports Clef and Clef Flash, Cloudflare's new open-source decision models, through /v1/systemone...
For 在智能体里做路由、意图识别或审核的工程团队、本地部署模型、用 Ollama 做推理的开发者、需要低延迟分类与打分的端侧应用
Sources
#The public sources behind the published signals in this briefing.
Subscribe
The public feeds carry published briefings only, never drafts or archived ones. They are in Chinese.