2026年10月22日 技术热点总结

项目简介：给国产电视的一块干净桌面：参照 tvOS 设计的开源 Android TV 桌面，没有广告，开机自动接管。A tvOS-style launcher for Xiaomi, TCL and other China-market TVs.

技术栈：Kotlin

核心介绍：参照 tvOS 设计的 Android TV 桌面：没有广告，没有推荐流，只有你的 App、你的壁纸和 Apple TV 航拍屏保。 为小米、TCL 等国产电视而做，开机自动接管，待机也不会被清理。

关键特性：时钟：大号细体时钟压在壁纸上，壁纸底部和左侧自动压暗。往下滚动时，时钟收进状态栏。；应用网格：每行 5 个，聚焦时浮起并显示名字。默认加一圈白色描边，离得远也看得清；在「控制中心 → 个性化 → 焦点样式」可以改回 tvOS 的原样。；图标：有电视横幅就用横幅；方形图标用自己的底色铺满磁贴；自适应图标拆成两层做视差。；全部 App：小米「应用中心」「高清播放器」「米家」这类没。

项目数据：⭐ 2,044 Stars，🍴 146 Forks

项目简介：REAP – Prompts for you to use REA (Reverse Engineer Anything), get the most out of it, and clone or break down any program

技术栈：prompt-engineering、rea、reverse-enginnering

核心介绍：A curated collection of prompts for REA (Reverse Engineer Anything) — the toolset that lets your AI coding agent inspect how software works, from native binaries, firmware and kernels to games, Electron apps, APKs and websites. Prompts are ready to paste into your agent. Replace everything in and keep each request focused on one feature — narrow questions get evidence-backed answers. (quoting Ma…

项目简介：Make an old Android TV fast again with Claude Code or Codex: a tiny ad-free home screen plus the prompt to clean it up

技术栈：Java

核心介绍：Claude Code or Codex does the work. You just press OK on the remote. 🧹 Removes the junk Bloatware, ads and sponsored rows, gone. 📺 A calm home screen Your apps, your HDMI inputs and a clock. Nothing else. ↩️ Undo anytime Apps are turned off, never deleted. Settings → Device Preferences → About → press Build 7 times. Then turn on USB debugging in Developer options. Stuck? Read the notes and warnings first. The laptop and TV must…

技术栈：JavaScript

核心介绍：Your TV should do what you paid for: your apps, your HDMI inputs, and nothing else. No ads from the TV itself, no watch tracking, no update you did not ask for. dumbtv applies that over ADB from another machine — no root, nothing written to /system, every step reversible with one script. The same idea, told to whichever TV will listen: Tested end to end on a Fire TV Edition television running Fire OS 7.7.1.7 (Android 9). Fire OS 6 and 8 differ in small ways; Android TV needs a different

项目数据：⭐ 1,156 Stars，🍴 97 Forks

Test-time scaling (TTS) improves large language model (LLM) reasoning by sampling multiple answers, but fixed sampling budgets ignore both query difficulty and the changing energy cost of inference under GPU contention. We measure up to a 4.1-fold increase in energy for the same four-sample query as GPU load rises. We present ATEST, a resource-aware TTS framework that jointly considers the expected value of additional samples and their current GPU cost. ATEST-Online, its default mode, requires no offline calibration: it incrementally generates samples, estimates the stability of the plurali…

The physical world has continuous state and action spaces, yet robot planning often benefits from hybrid discrete-continuous formulations that explicitly expose combinatorial structure, such as contact modes or discrete primitive actions. Existing methods typically rely on human-defined symbol sets or symbols derived from language supervision, without asking what makes a symbol computationally useful. In this paper, we take an inference-algorithm–centric view to define usefulness: a good symbol should expose the discrete commitments that simplify solving the original planning problem. We f…

A fine-tuned Vision-Language-Action (VLA) policy can easily place the alphabet soup in the basket on command, just as it saw during fine-tuning. However, if the basket is shifted five centimeters to the left, it drops the soup off the table. We developed an interpretability pipeline to investigate what such a policy is computing. Injecting task A's activations into the action expert while the policy looks at task B's scene makes pi0.5 execute task A's motor trajectory in 99.6% of episodes (n{=}1{,}968). The injected program is bound to absolute workspace coordinates and ignores the visible…

Bimanual manipulation with dexterous hands has attracted growing attention in research on robotic action models. However, few studies have developed benchmarks specifically tailored for dexterous manipulation, leaving the dexterous manipulation capabilities of existing models insufficiently characterized. In this paper, we present BiDex-100, a simulation benchmark for bimanual dexterous manipulation that combines broad task coverage, multi-dimensional evaluation, and high-DoF hand embodiments. Concretely, the benchmark comprises 106 tasks organized along five capability axes: generalization…

Training coding agents typically updates the policy while keeping the harness fixed. During training, trajectories reveal flawed repair strategies and interaction problems that vary across tasks and change as the policy learns. We introduce HarnessAdapt-RL, which trains the policy and adapts its harness during reinforcement learning. For flawed repair strategies, a frozen teacher generates task-specific mechanism-level skills. For interaction problems recurring across tasks, a separate teacher model evolves tools and hooks, keeping each change only if it improves the observed pass rate or t…

Recent research-level physics-reasoning benchmarks for LLMs reveal a failure mode that lies not in knowledge or execution, but in modeling the world being reasoned about. Models often silently alter given quantities, fabricate laws, or inject unstated assumptions to force a problem into familiar patterns, and then reason inside this corrupted world. We define this premise-level divergence as modeling drift: the gap between a model's internal axiomatic world model (AWM) and the true physical world specified in the problem. Although RLVR and PRM have yielded substantial gains in mathematics a…