2026年10月27日 技术热点总结

项目简介：关掉国产安卓手机系统里的广告、热搜、资讯推荐和营销推送。不 root，可还原，支持 AI 智能体一键执行。

技术栈：Python、adb、ai-agent、android、coloros、debloat、hyperos、magicos、originos

核心介绍：关掉国产安卓手机系统里的广告、热搜、资讯推荐和营销推送。 介绍页：https://rddmwmddr.github.io/android-ad-cleaner/ 插上手机，点几下按钮。不需要 Python，也不需要 AI 工具。 > v0.2.3 · 2026-10-25：处理读屏服务 already registered 冲突：退出前释放本次服务，同机任务互斥，短暂冲突只重试一次；持续占用时给出恢复步骤并跳过后续读屏任务。收尾不再误报全部完成。保留通知扫描、完整诊断及此前修复。旧版不会自动更新，请重新下载，已有备份会保留。详见 更新记录。 > 最新源码补充（尚未进入 0.2.3 安装包）：新增一加 Ace 2 Pro / ColorOS 14 的实测与有限适配，修复独立通知设置页识别、混合付款通知排除和应用启动后退出到上一应用的保护。范围与未完成项见 一加旧系统实测。 > 2026-10-26 源码更新（未发布安装包）：读屏失败时提示排查 GKD…

项目简介：词光 · Word Atlas — Illustrated IELTS vocabulary with 7,298 words, UK/US audio and FSRS review. Open-source app + CC BY 4.0 dataset.

技术栈：TypeScript、chinese、dataset、english-learning、flashcards、fsrs、ielts、language-learning、pwa

核心介绍：7,298 个雅思词条 · 原创插画 · 英美发音 · FSRS 间隔复习 立即体验 ↗　 /　 48 秒介绍　 /　 下载词库　 /　 English 词光是一款以插画和语境为核心的雅思背词应用。看画面理解词义，听发音建立联系，再用主动回忆和间隔复习巩固记忆。无需账号，学习进度保存在自己的设备上。 > 版本说明：这个仓库提供词光的开源基础版及配套数据集。在线网站独立迭代，功能与仓库可能不同；仓库里的代码和内容分别遵循下方的 MIT / CC BY 4.0 授权。 产品预览 · 学习方式 · 本地运行 · 开放数据 · 参与贡献 每日计划，把注意力放在今天插画、词义、发音，在一张词卡里 以上为仓库开源版的实际界面。页首为品牌概念插画。 19 个话题组成词汇星图。拖动旋…

项目简介：The agent skills I use every day. Works with Claude Code and Codex.

技术栈：Python

核心介绍：The agent skills I use every day, cleaned up so anyone can use them. They work in Claude Code and Codex. Some use T3 Code features when they are available and fall back to native tools when they are Some skills call others. Install the ones they need too:

项目数据：⭐ 1,541 Stars，🍴 83 Forks

项目简介：Steering-first neuron explorer for Qwen3.5-4B: push every MLP neuron, keep every rollout, browse the results

技术栈：Python、interpretability、mechanistic-interpretability、qwen、steering

核心介绍：Steering-first neuron explorer for Qwen3.5-4B. Every MLP neuron in a layer is pushed up and down on a few prompts, every rollout is kept, and the explorer lays the results out over a 3D view of the model. uv run python explorer/server.py # http://localhost:8791 Reads sweep runs from runs/ and per-layer stats from stats/, both produced by the sweep below; the caches in explorer/cache/…

项目简介：Agent Skills that teach coding agents to write, review, and port idiomatic Rust, plus a tested clippy config. Works with Claude Code, Codex, Copilot, Cursor, and Gemini CLI.

技术栈：Python、agent-skills、claude-code、claude-plugin、claude-skills、clippy、code-review、coding-agent、cursor

核心介绍：Three standard Agent Skills and a lints config that teach coding agents to write, review, port, and enforce idiomatic Rust: toward idiomatic, robust, maintainable code. prioritized report where every finding cites the rule it breaks and shows the fix. behavior, porting the original tests first s…

Biometric authentication systems increasingly use privacy-preserving mechanisms to protect sensitive templates. However, the authentication interface itself may still leak information that enables reconstruction attacks. Prior attacks exploit released similarity scores, whereas more advanced systems reveal only the binary accept/reject decision. This binary feedback provides substantially less information, making template reconstruction far more challenging.
In this work, we introduce query-efficient reconstruction attacks under extremely limited feedback, addressing two key limitations o…

When humans collaborate with AI agents on long-horizon tasks, what happens to the agents' contributions? For solving long-horizon tasks, humans place greater importance on ongoing interaction with AI agents (e.g., revising plans) than on an agent's ability to complete the task in a single attempt. However, most agentic benchmarks have focused more on automation. Strong performance on such benchmarks does not imply effective real-world collaboration. Professional workflows offer a useful setting, as they often involve interdependent steps. To address this gap, we introduce Kolab, a new datas…

In joint-embedding predictive architectures (JEPA), an encoder and a predictor are trained together to produce a representation that minimizes the error in predicting the representation of the next observation. Because the representations of both the current and the next observation are learned, it is unclear which features they will include. Earlier characterizations describe the features JEPA selects in terms of their individual predictability, raising the concern that unpredictable features, like a traffic light, will not be encoded into the representation. However, the prediction error …

Generative reward models (GenRMs) generate natural language rationales to justify preference judgments, yet a correct verdict does not guarantee a sound justification. Existing approaches often rely on reference explanations to supervise rationales, but obtaining such references can be costly and does not necessarily establish whether generated rationales support the resulting verdict. To address these limitations, we propose DeCert, a framework that evaluates the validity and verdict consistency of generated rationales without requiring reference explanations. Crucially, DeCert translates …

Recent methods for injecting graph structure into LLM prompts include graph soft prompting, in which a graph neural network (GNN) learns graph-specific tokens that capture relational information among entities in textual prompts. G-Retriever combines graph soft prompts with a verbalized description of the same graph to strengthen the textual prompt signal. We conduct an in-depth analysis of G-Retriever and uncover a pathological behavior we term token collapse. Across three benchmarks, three graph encoders, two LLM families, and both frozen and LoRA-adapted models, G-Retriever's graph token…

The life-cycle of robotics foundation models crucially depends on data labels. Outcome reward labels are needed for monitoring real-world deployments. Process reward labels are needed during RL post-training. Timestamped captions are needed for filtering and balancing pretraining data. We present Flora, an industrial-quality data engine used for training GR00T models, that simultaneously covers these capabilities. Flora achieves SOTA scores on the captioning benchmark WGO-Bench and reward benchmarks RoboReward and Robometer. On internal benchmarks it is competitive with GPT-6 Astra while be…