基于来源的发现
新词雷达
发现公开来源中出现的新名称与概念,查看带日期的证据、来源覆盖和可比较增长。
收藏仅保存在当前浏览器,不会自动同步到账户。关闭本页面不会停止服务器的数据采集。
采集覆盖 · 历史不足本期 / 前期观测天数 7 / 4
统计使用完整 UTC 日。“首次观测”指本站首次收录,不代表词语的诞生时间。
本期窗口:2026-10-30 → 2026-11-06 (不含结束日) · 历史跨度 12 天
| 来源 | 本期天数 | 前期天数 | 最近成功采集 |
|---|---|---|---|
| arXiv | 6 | 4 | 2026年11月6日 00:30 UTC |
| arXiv | 7 | 3 | 2026年11月6日 05:15 UTC |
| Ask HN | 7 | 4 | 2026年11月6日 15:45 UTC |
| Ask HN | 7 | 4 | 2026年11月6日 15:30 UTC |
| Ask HN | 7 | 4 | 2026年11月6日 16:00 UTC |
| Ask HN | 7 | 4 | 2026年11月6日 16:00 UTC |
| GitHub | 7 | 4 | 2026年11月6日 12:45 UTC |
| GitHub Issues | 7 | 4 | 2026年11月6日 16:00 UTC |
| GitHub Releases | 7 | 2 | 2026年11月6日 12:15 UTC |
| GitHub | 7 | 4 | 2026年11月6日 14:30 UTC |
| GitHub Issues | 7 | 4 | 2026年11月6日 16:45 UTC |
| GitHub Releases | 7 | 4 | 2026年11月6日 16:45 UTC |
| GitHub | 7 | 4 | 2026年11月6日 14:30 UTC |
| GitHub Issues | 7 | 4 | 2026年11月6日 17:00 UTC |
| GitHub Releases | 7 | 4 | 2026年11月6日 16:45 UTC |
| GitHub Discovery | 1 | 0 | 2026年11月6日 16:00 UTC |
| GitHub Trending | 7 | 4 | 2026年11月6日 13:15 UTC |
| 7 | 3 | 2026年11月6日 08:30 UTC | |
| Hugging Face | 7 | 3 | 2026年11月6日 08:30 UTC |
| Hugging Face Spaces | 1 | 0 | 2026年11月6日 15:30 UTC |
| Hacker News | 7 | 4 | 2026年11月6日 13:15 UTC |
| Hacker News | 7 | 4 | 2026年11月6日 16:15 UTC |
| Hacker News | 7 | 4 | 2026年11月6日 16:15 UTC |
| Hacker News | 1 | 0 | 2026年11月6日 16:00 UTC |
| Show HN | 1 | 0 | 2026年11月6日 15:45 UTC |
| MIT News | 7 | 3 | 2026年11月5日 16:15 UTC |
| OpenAI | 7 | 4 | 2026年11月6日 11:45 UTC |
| Product Hunt | 7 | 4 | 2026年11月6日 13:15 UTC |
| 量子位 | 6 | 4 | 2026年11月4日 10:15 UTC |
| OpenRouter | 1 | 0 | 2026年11月6日 17:03 UTC |
| Hugging Face Models | 1 | 0 | 2026年11月6日 17:03 UTC |
| HF Daily Papers | 1 | 0 | 2026年11月6日 17:03 UTC |
| OpenRouter | 1 | 0 | 2026年11月6日 17:03 UTC |
| SEC EDGAR | 7 | 4 | 2026年11月6日 10:45 UTC |
| Sequoia Capital | 7 | 3 | 2026年11月6日 06:45 UTC |
| Simon Willison | 1 | 0 | 2026年11月6日 15:15 UTC |
| TechCrunch | 7 | 4 | 2026年11月6日 15:00 UTC |
| Contracts Finder | 7 | 4 | 2026年11月6日 08:00 UTC |
| V2EX | 7 | 4 | 2026年11月6日 15:00 UTC |
- 计数仅代表已监测的公开来源,不代表全网热度或搜索需求。
- 两周均需足够且可比较的采集覆盖,才判断增长。
七日指标使用上方完整 UTC 窗口。优先级用于安排研究顺序,不代表搜索量、市场规模或成功概率。
RefactorProof
arXiv:2610.36328v1 Announce Type: new Abstract: Benchmarks for verifiable code generation usually test a proof against one fixed implementation. We ask whether the same proof remains valid after a program refactoring that preserves behavior. We introduce RefactorProof, which derives 458 Lean-certified variants from the 185 VERINA tasks that compile in our pi
首次观测 2026年10月28日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 14 / 100
SPY
arXiv:2610.35925v1 Announce Type: new Abstract: Training contextual masking systems for personally identifiable information (PII) requires examples distinguishing maskable from preservable mentions of one entity type, but privacy restrictions limit access to real data. Author-based augmentation (SPY) labels mentions unrelated to the author as non-PII, presum
首次观测 2026年10月28日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 14 / 100
IdoGol24/weir
Hi HN. Weir is for one problem: credentials crossing the boundaries of your AI agent. For agents you build, Weir plants fake credentials where real ones would live, runs the agent's normal flows, and checks what comes out: telemetry, logs, console output, model requests, and requests to other services. It looks for those exact values, including common encode
首次观测 2026年10月27日 · Hacker News
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 14 / 100
BehaveRec
arXiv:2610.36239v1 Announce Type: new Abstract: Sequential recommendation models typically predict a user's next action from recent session interactions, which limits personalization when short-term signals are weak or ambiguous. Rather than assuming that incorporating longer-term history and community signals is universally beneficial, we study when it actu
首次观测 2026年10月28日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 14 / 100
-
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
Hackerscad
首次观测 2026年10月28日 · Hacker News
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
DELA
arXiv:2610.32096v1 Announce Type: new Abstract: Diffusion language models (DLMs) support bidirectional denoising and iterative refinement, but generation remains largely controlled by token-level masking schedules. This leaves draft-level refinement implicit and limits explicit control over when to insert missing content, delete redundant spans, revisit unce
首次观测 2026年10月26日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
界目前比较主流
目前感觉,应该是不需要一个专门的语音输入法,只需要一个快捷键能调起它,或者有其他方式能调起它就行了,对不对? 我搜了一下,发现开源界目前比较主流的就是千问的那种语音模型,对不对?那这样的一个语音模型在本地运行,外加 AI 润色,这样就成了,是不是?它的优点就是 ASR 的流式输出? Typeless 看起来很好,是收费的。我习惯于长按快捷键输出,它只有按一下快捷键进行转换的方式,也不太合适。 我搜了一下,有一个叫 FluidVoice 的,还有一个叫 Handy 的,好像都不错。FluidVoice 据说是针对性优化做得最好,咱也不懂。FluidVoice 不支持更多的中文模型,主要还是面向西方社区。Handy 据说是用 Rust 写的,但还有 WebView 套壳。好家伙,这一下子感觉 Rust 就白用了。
首次观测 2026年10月28日 · V2EX
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
Formstep
I built Formstep: https://formstep.io The idea came from a problem I kept running into while working on loan applications. A workflow could automate a large part of the process, but eventually it would need something only the customer could provide: a document, a correction, a signature, an explanation, or confirmation of some information. At that point, wor
首次观测 2026年10月27日 · Hacker News
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
multica-ai/andrej-karpathy-skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
首次观测 2026年10月26日 · GitHub Trending
- 七日提及
- 0 前期 0
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
-
- 七日提及
- 0 前期 0
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
Clivern/Sloth
首次观测 2026年10月26日 · Hacker News
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
-
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
Personality Latents
arXiv:2610.36805v1 Announce Type: new Abstract: Controlling the personality of a language model usually relies on written profiles or named trait categories, so it can only express what an author writes down or what a taxonomy names. We instead propose Personality Latents (PL), which learn a latent for each person from their own writing, with no personality
首次观测 2026年10月28日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
地图下载器补了历史影像
我是 GeoD 的开发者,最近把 Google Earth 历史影像接进了桌面版的 Wayback 。 这次主要想把一个流程改顺:看历史影像时,不用先框选、查完再手动加载。选好日期以后,拖动、缩放地图,影像会跟着视野加载;找到需要的资料后,再选范围导出。 实现上用了 Keyhole 方案,本机处理历史目录和影像读取,再交给地图层。日期切换、在途请求和缓存分别处理,避免快速切换时串图。 下载也补了准备阶段的进度反馈。之前有的小任务点下去没动静,过一会儿直接完成;现在能看到准备索引、读取和导出,文件写完才显示完成。 另外一起改了几个小地方: - 历史图源集中在 Wayback ,新增 NASA Landsat 、EOX 年度影像及星图授权图源适配。 - 区域选择放回顶部,日期保留一个入口。 - 时间轴点击范围扩大,
首次观测 2026年10月28日 · V2EX
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
PTQ
arXiv:2610.32127v1 Announce Type: new Abstract: Post-training quantization (PTQ) is a key technique for efficiently deploying large language models, as it compresses weights into low-bit representations without retraining. Existing PTQ solvers, including GPTQ, LDLQ, and block coordinate descent, commit each weight block to a single codeword before its neighb
首次观测 2026年10月26日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
affaan-m/ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
首次观测 2026年10月26日 · GitHub Trending
- 七日提及
- 0 前期 0
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
-
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
-
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
NVIDIA/SkillSpector
Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, security risks, prompt injection, data exfiltration, and supply-chain risks in Claude Code, Codex, and MCP skills before you install them.
首次观测 2026年10月26日 · GitHub Trending
- 七日提及
- 0 前期 0
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
CONCUR
arXiv:2610.36437v1 Announce Type: new Abstract: When a Large Language Model (LLM) is trained with Reinforcement Learning (RL) on multiple tasks, a task's primary reward can gate auxiliary rewards such as a length reward: on a failing response, an auxiliary cannot exceed the primary. However, group-relative advantage estimators treat a gated auxiliary like an
首次观测 2026年10月28日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
Agent Handshake by Fini
Agent-to-agent customer support that holds its answer
首次观测 2026年10月21日 · Product Hunt
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
实现
这段时间按照 DeepSeek 的 DSec 论文,做了一个 sandbox platform 的开源、自托管实现:OpenDsec。 项目地址: https://github.com/Kernflux/opendsec 先说明一下:这是独立项目,不包含 DeepSeek 代码,也不是 DeepSeek 官方发布或维护的项目。架构和主要设计来自论文;论文没有明确说明、或者我们采用不同方案的地方,都在 SPEC 里单独标注了。 整体架构分为控制面和节点运行时。 控制面里,stateless apiserver 负责认证、配额、调度和请求转发;IAM 用 Raft 保存身份、项目和配额状态;watcher 汇总各节点的健康与负载;每台计算节点运行一个 edge,负责最终准入和 sandbox 生命周期。组件之间走
首次观测 2026年10月28日 · V2EX
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
OCR
arXiv:2610.33004v1 Announce Type: new Abstract: Optical Character Recognition (OCR) is essential for digitizing historical archives, yet standard evaluation metrics such as Character Error Rate (CER) do not necessarily reflect whether meaningful information, such as names, dates, and identifiers, is preserved. In this work, we investigate entitylevel analysi
首次观测 2026年10月27日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
-
- 七日提及
- 0 前期 0
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
Cryo
Cryo is a programming language that models long-running tasks like workflows or AI agents. Programs freeze to disk when idle and can be thawed when new data arrives. As part of our announcement of Endothermic Labs, we've put together a small demo of Cryo that highlights its features and walks through the language and its plugin system. We originally built it
首次观测 2026年10月27日 · Hacker News
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
anomalyco/opencode
The open source coding agent.
首次观测 2026年10月11日 · Hacker News / GitHub Trending / TechCrunch
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
-
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
DocSkill
arXiv:2610.36759v1 Announce Type: new Abstract: Multilingual multi-hop document question answering (MMHop-DocQA) requires locating and integrating evidence across complex layouts, document pages, and diverse languages. Existing approaches include single-pass processing and adaptive tool use, but learning to coordinate evidence acquisition, cross-page integra
首次观测 2026年10月28日 · arXiv
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100
RandComp Studio
首次观测 2026年10月27日 · Hacker News
- 七日提及
- 0 前期 1
- 七日来源
- 0
- 七日增长
- 不可比较
- 优先级
- 13 / 100