广场

miniCOIL EN-ES: Sparse Neural Retrieval Across the Language Barrier

Sparse neural retrieval finally started getting more and more attention ( SPARSEUP , MILCO , sparse encoders in Sentence Transformers …). Well, amazing, we like attention! Our miniCOIL v1 sparse neural retriever article ended with a promise: to keep working on this sparse neural retriever in-depth, improving the model’s quality, and in-width, “extending it to other dense encoders and to languages beyond English.” Here is the next “in-width” step of the miniCOIL saga: to try and cross the language barrier, that is, make the model suitable for multilingual retrieval. It’s a fun challenge, as sparse retrievers are hard to adapt to this scenario: exact matching in cross-lingual text search usual…

QdrantAI 工具英文

The PSF end-of-year fundraiser

The PSF end-of-year fundraiser With my Python Software Foundation board director hat on: we've launched our end-of-year fundraiser for the non-profit that supports Python development and the Python community. If your company or your work benefits from Python, please consider donating this year! The PSF looks after the IP behind Python, operates PyPI, sponsors core developer sprints, supports community events (including running PyCon US every year), and employs a small and excellent staff team to keep everything running. We also operate the Developers in Residence program , which now includes three security developers (welcome, Brett Smith) focused on helping secure Python, PyPI, and the wide…

OpenAI 发布 GPT-6.2 Astra 和 Sol;王兴兴:家用机器人或将迎来技术突破;《GTA 6》开发商批评 AI 公司放任游戏仿制行为|极客早知道

OpenAI 发布 GPT-6.2 Astra 和 Sol,强化编程、电脑操作与专业工作 OpenAI 正式发布 GPT-6.2 Astra 和 GPT-6.2 Sol 两款模型,重点提升专业任务、Coding、电脑操作和执行效率,并降低部分使用成本。新模型将应用于 ChatGPT Work、Codex 和 API。 其中,Astra 面向复杂编程和专业工作,Sol 则更强调日常多模态任务及工具使用。 OpenAI 同时发布系统安全卡,披露两款模型在网络安全能力方面均达到其 Preparedness Framework 的 Critical 门槛,Astra 的 AI 自我改进能力达到 High 门槛,因此加强了高风险能力的访问限制。(来源:OpenAI) 谷歌 DeepMind 首席战略官跳槽 Anthropic,AI 人才争夺持续升温 11 月 17 日消息,据《The Information》报道,谷歌 DeepMind 首席战略官贾斯吉特·塞孔(Jasjit Sekhon)已离职,正式加入 Claude 开发商 Anthropic。 塞孔在社交平台 X 上确认了这一消息,表示未来将致力于提升 Claude 的知识可靠性与事实严谨性。他今年 3 月加入 DeepMind,任职仅约 8 个月。此前曾在哈佛大学、加州大学伯克利分校及耶鲁大学任教,并担任桥水基金高级科学家兼 AI 负责人。 今年以来,谷歌 DeepMind 已有多名核心人才出走。6 月,Gemini 联合负责人诺姆·沙泽尔转投 OpenAI;参与开发 AlphaFold 的诺贝尔化学奖得主约翰·江珀也加入 An…

极客公园AI 工具中文

Why Hyperbolic Embeddings Matter for Hierarchical Data

We choose embedding models, dimensions, and indexes. The geometry usually comes with the package. But why use a flat space, and what else could we choose? What Is a Manifold, and Where Do Our Vectors Live? A manifold is the space our embeddings live in. For embeddings, we care about the geometry we give that space. It determines how we measure distance, what the shortest path looks like, and how much room there is as we move outward.

QdrantAI 工具英文

External sandboxes for coding agents

External sandboxes for coding agents I've been exploring new ways to sandbox code that is executed by agents, with a focus on making it safer to try out local models without worrying that a mistake or prompt injection attack might cause them to take damaging actions on a machine. I worked on this project with Prime Radiant , the AI research lab behind Superpowers and other projects . My post on the Prime Radiant blog introduces the external sandbox pattern, where code executed by the agent runs inside a dedicated sandbox while the agent itself lives outside of that sandbox. To demonstrate this pattern we've released three sandbox server implementations: external-sandbox-v86 runs a Node.js se…

最大规模具身大模型挑战赛落幕,48 小时真机总决赛比出了什么?

开发者围绕同一块基础模型进行实验,让模型直面成功的方法、失败的经验和真实世界的问题。 作者|Li Yuan 编辑| 郑玄 最大规模具身大模型挑战赛总决赛结果出炉,排名第三的队伍 SST 只额外采集了 13 条数据。 他们没有大量补数据,而是让 DeepSeek 先理解任务,再用视觉分割模型标出目标物体,最后交给 LingBot-VLA 2.0 完成动作,并在真机评测中拿到 65.87 分。 通用大模型与具身模型的协同在这里变得很具体:DeepSeek 负责弄清楚要做什么,具身基础模型帮机器人找准操作对象、做出动作。 11 月 16 日,首届蚂蚁灵波具身大模型挑战赛在上海张江收官。经过线上初赛,15 支战队进入为期 48 小时的线下总决赛。所有队伍使用同一款开源模型 LingBot-VLA 2.0、同一款松灵 COBOT MAGIC 机器人和统一的现场算力,完成蔬果摆放、化妆品收纳和双臂清扫三项任务。 起点相同,解法不限:有人重采数据,有人调整模型的训练和推理方式,也有人接入通用大模型、视觉分割或仿真工具。48 小时里,15 支队伍要回答的是同一个问题:面对真实桌面上的物体,怎样让已有的模型把这件事做成? 01 当胜负开始取决于如何使用数据 纵观整场比赛,有一点十分有趣:几乎每支队伍都提到了数据工程的重要性。 LingBot-VLA 2.0 已经具备基本的视觉理解、指令理解和动作生成能力。参赛者不必在 48 小时里重新训练一个机器人大脑,而可以直接判断模型已经会什么、会在哪些情况下失败,再决定应该补什么数据、怎样训练。 冠军团队 TARS 首先给三项任务分出了难易。经过测试,他们…

极客公园AI 工具中文