2026年6月28日 技术热点总结

项目简介：DeepSpec: a full-stack codebase for training and evaluating speculative decoding algorithms

技术栈：Python

核心介绍：DeepSpec is a full-stack codebase for training and evaluating draft models for speculative decoding. It contains data preparation utilities, draft model implementations, training code, and evaluation scripts. Install the Python dependencies: python -m pip install -r requirements.txt Data preparation additionally requires an inference engine to serve the target model when regenerating answers; see scripts/data/README.md for details. Run the stages in order — each sta…

项目简介：A single archive of public exploit PoCs and vulnerability research writeups. At the time I post these, none have been reported. Feel free to report them yourself and take credit for the CVE if handed out lulz. Please do not abuse these. I do this so to allure people into the field, and I’ve always found this is the most efficient way.

技术栈：Python

核心介绍：This repo was incomplete when published. That’s why some findings are kinda ass (ghidra) and some are better. Going forward, only serious vulnerabilities will be shared (Floci, libssh2, FFmpeg, c-ares). I’d also like to credi…

项目简介：A tiny macOS menu bar status indicator for Claude Code: animated icons, elapsed timer, and open/close lifecycle.

技术栈：Swift、anthropic、app、claude、claude-code、claude-desktop、cli、indicator、macos

核心介绍：A tiny macOS menu bar app that shows Claude Code’s live status: an animated Claude icon while it’s thinking or running a tool, a yellow dot when it’s awaiting your permission, and the elapsed time of the current turn. Lightweight, no window, no dock icon, no usage dashboards. > Built so you can tab away during a long “thinking” stretch and still see, at a glance, whether Clau…

项目简介：这是一个基于tabbit的研究包，可以转化成OAI格式出来，同时增加了会员认证功能和一键提取cookie的浏览器拓展，方便快速本地快速使用claude gpt等模型

技术栈：JavaScript

核心介绍：Tabbit 是一款基于 Chromium 的国产 AI 浏览器,内置了 21 个 AI 模型(Claude-Opus-4.8、GPT-5.5、Gemini-3.5-Flash、DeepSeek-V4-Pro 等)。正常情况下你必须打开 Tabbit 浏览器才能用这些模型。 你的客户端 ──OpenAI格式──▶ Tabbit2API(本地服务) ──翻译+签名──▶ web.tabbit.ai 1. 装了 Tabbit 浏览器并能登录账号(tabbit.ai 下载) 2. Pro 会员:在 Tabbit 设置里”设为默认浏览器”可解锁(免费),premium 模型(Claude/GPT/Gemini)必需 3. Node.js 18+(本项目零依赖,无需 npm install) Tabbit 的登录态存在 Cookie 里(含一个 HttpOnly 的 JWT token),需要从浏览器导出一次。 1. 打开 Chrome 或 Tabbit 浏览器,地址栏输入 chrome://extensions/…

项目简介：The cozy life-sim that pays. Build a home, master skills, climb careers — earn $PLUM and cash out real SOL. Free to play in your browser.

技术栈：JavaScript

核心介绍：Plumtown is a cozy, browser-based play-to-earn life simulator. Raise a Sim, keep them happy, build a home, climb a career, fall in love, start a family — and every real milestone you hit mints $PLUM you can withdraw on-chain. Built from scratch with vanilla JavaScript, HTML & CSS — no frameworks, no build step, no front-end dependencies. CA: 5jXuVv7KfWmjCf7PAB1cNXbNieP1fno6eWZDVBLppump

项目数据：⭐ 354 Sta…

项目简介：A Rust toolkit for the Oura Ring (Gen 3/4/5): reverse-engineered BLE protocol, event decoders, and reimplemented data-processing algorithms. Sync, store, and analyze your data locally.

技术栈：Rust

核心介绍：Reverse-engineering the Oura ring BLE protocol, plus an independent, cloud-free client that reads your data straight from the ring. Tested live against a Ring 3 Horizon and a Ring 5 (pairing, auth, and event sync confirmed on both). Designed for Ring 3/4/5, which share the same GATT layout, packet framing, and authentication flow. Straight from the ring, with no Oura accou…

项目简介：基于花生十三公开教学体系、课程资料、历年正题整理得到的skill。

核心介绍：> *”备考路上，技巧为王，坚持为皇！”* 行测做不完？申论写不出？数量关系全蒙C？ 别慌，你只是缺一套系统化的解题武器库。 快速开始 · 模块速查 · 答题策略 · 效果示例 当你想说这些话的时候，直接开口： 直接发题目、问方法、要计划，不用记任何命令。 > 场景一：题目讲解 用户 ❯ 缔约过失责任是指在合同订立过程中，一方因违背依据诚实信用原则所产生的义务，而致另一方的信赖利益受损失，就应承担损害赔偿责任。 根据以上定义，下列选项中乙方不可以要求甲方负缔约过失责任的是（　　）。 甲、乙公司经过多轮磋商对机床价格达成了一致，于是乙公司派人来与甲公司签约，甲公司突然加价，乙公司未同意，甲公司遂拒绝签约，乙公司因此不能及时开工 甲公司得知乙公司正在就贸易合同与丙公司谈判，为了排挤丙公司，就向乙公司提出了更优惠的条件。丙公司退出后，甲公司借故终止谈判，乙公司因此受损 甲厂与乙厂达成了药材供应协议，约定每月向乙厂供药材10吨，后药材价格暴涨，甲厂希望能重新商定价格，遂停止供货，乙厂因此失去大批客户

项目数据：⭐ 350 Stars，🍴 69 Forks

ABACUS is a unified vision-language model that handles object counting, crowd counting, referring-expression counting, and count-faithful image generation without any benchmark-specific training required. Our model is built on existing 3B-parameter unified foundation model and is adapted for object localization tasks using three key innovations: density-aware adaptive zooming with objectness maps for spatial grounding; a boundary-aware count policy via GRPO to eliminate crop-boundary errors; and a cycle-consistent GRPO strategy where the understanding branch self-critiques generated outputs…

Process reward models enable fine-grained, step-level evaluation of LLMs, yet building them for agentic settings remains prohibitively difficult: long-horizon interactions, irreversible actions, and stochastic environment feedback make both human annotation and Monte Carlo estimation infeasible at scale. In this work, we show that reinforcement learning (RL) post-training already provides the ingredients for effective step-level scoring, eliminating the need for dedicated reward model training altogether. Concretely, we derive an implicit advantage under a general stochastic Markov decision…

Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression methods mainly rely on attention weights to estimate token importance. While attention effectively captures contextual relevance, it overlooks complementary information-theoretic signals related to predictive uncertainty and token informativeness. In this paper, we revisit token importance from a forward-looking perspective and introduce Forward Influence, a metric that measures how compressed t…

Earth Observation (EO) forecasting aims to predict future Earth surface dynamics from satellite observations under changing meteorological conditions. In this paper, we view this task as a partially observed, weather-driven world modeling problem, in which weather acts as a conditioning signal, while forecasting remains uncertain due to sparse observations and unobserved land-surface states. However, existing methods do not fully capture this setting: deterministic models collapse uncertainty into a single future prediction, while diffusion-based methods typically treat weather variables as…

The prevalent dual-branch paradigm, i.e., training a side network to encode visual conditions and fusing its intermediate-layer features to a frozen pretrained main network, has shown remarkable success in visual-condition controllable generation. Despite its widespread adoption, the role of the side branch and its training efficiency remain underexplored. In this paper, we first revisit this mainstream paradigm through the lens of score-based generative modeling: 1) The main network preserves visual perceptual quality by providing a prior unconditional score. 2) The side network steers con…

As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities while overlooking broader dimensions, resulting in saturated performance on modern agents and failing to probe their limitations. To this end, we introduce GauntletBench, a web-based benchmark for evaluating agent generalisation in challenging scenarios, focusing on three underexplored capabilitie…