2026年8月6日 技术热点总结

项目简介：让 AI 写的中文读起来像一个具体的人在说话。通用创作与改稿 Skill，开箱即用。

技术栈：Python、agent-skills、chinese-writing、creative-writing、writing-skill

核心介绍：快速安装 · 写作流程 · 仓库结构 · 提交问题 > AI 写中文有个通病：读完觉得挺流畅，但说不出是谁写的。活人感写作想治的就是这件事。 让模型写出来的文章读起来像一个具体的人在说话——知道一些事，有判断，偶尔岔开一句，还能接回来。适用于知乎回答、公众号文章、博客、论坛帖、人物故事、科普、评测、小说、口播等大多数中文写作场景。 写作之前先解决一个前置问题：你手上有没有东西可写。 现实题材，材料不够就去查，查不到就追问或者缩短篇幅，绝不拿车轱辘话凑字数。虚构题材可以自由创造人物和情节，但每个场景仍然要有目标、有动作、有变化。 材料过关之后管三件事： 初稿写完还有一道关。Skill 会逐段检查有没有在原地转圈，砍掉重复解释，调整长短句节奏，拦住冒号滥用、破折号、「不是……而是……」之类的翻案腔和常见 AI 黑话。检查脚本只管已经写明的硬规则，不替你决定风格。

项目数据：⭐ 1,022 Stars，🍴 99 Forks

项目简介：RealReplicaBench: Benchmarking Long-Horizon Agents in High-Fidelity, Stateful, and Reproducible Replicas of Real Online Services

技术栈：HTML

核心介绍：Developed and maintained by the Accio team at Alibaba International. Overview · Live leaderboard · Mock showcase · Quick start · Reproducibility · Contact We run models on re

项目数据：⭐ 1,018 Stars，🍴 69 Forks

项目简介：Any icon morphs into any other — universal morphing for stroke-based icons with spring physics. Zero dependencies, ~7 KB gzip.

技术栈：TypeScript、animation、icons、morphing、react、svg、typescript

核心介绍：morphicons morphicons.com — live playground Universal morphing for stroke-based icons (Lucide, Tab ler, Heroicons, Iconoir, or your own paths): any icon morphs into any other with spring physics. Rotations are never declared by hand — they emerge from the math (2D Procrustes + polar interpolation). Zero runtime dependencies. import { MorphIcon } from “morphicons/react”; import…

项目简介：非官方 Kimi Slides Skill：让 AI Agent 生成可编辑 PPTD + PPTX，并附带本地浏览器编辑器 Unofficial Kimi Slides skill for AI agents — generate editable PPTD + PPTX with a local browser editor

技术栈：Python

核心介绍：简体中文 | English 逆向 Kimi Slides 实现的非官方演示文稿 Skill：让 AI Coding Agent 能够创建、编辑、复刻、读取并导出 PPT/PPTX。每次生成默认同时交付两份成果：可继续编辑的 PPTD 项目和开箱即用的 PPTX 成品（自动嵌入字体、写入淡入淡出切换动画），并提供本地在线 PPTD 编辑器支持随时手动导出。支持 Codex、Claude Code、Cursor、WorkBuddy 等任何兼容 SKILL.md 规范的 Agent。 > 本项目通过逆向分析 Kimi Slides Skill、PPTD 格式以及公开网页编辑器的前端行为与通信协议实现，并非 Kimi 或 Moonshot AI 的官方项目，也未获得其认可或支持。项目依赖的公开前端资源和兼容协议可能随 Kimi 更新而失效，仅供学习与研究使用。

**项目数据：*…

Reasoning language models frequently overthink: generating extended chains of behaviors such as hedging, approach abandonment, and self contradiction that consume tokens without improving answers. We show that these behaviors are not merely a consequence of length; even when controlling for response length, incorrect traces exhibit higher rates of unproductive self-reflection than correct ones. Addressing this requires identifying where self-reflection helps vs hurts, but obtaining these step-level annotations is costly. We observe that intermediate answer commitments within reasoning trace…

Self-consistency assumes the most frequent answer among sampled reasoning traces is the most reliable, but this can fail in causal reasoning: samples often repeat the same confounding error, and votes fragment across multiple valid answers, letting an invalid answer win despite a valid minority trace. We introduce CALVER (Causal Axiom-Level VERification), a training-free symbolic verifier that scores structured traces against Pearl's causal criteria, including -separation, backdoor adjustment, and intervention, and selects the highest-scoring candidate without consulting a reference answer….

Weight-only quantization substantially reduces the storage of large language model (LLM) transformer blocks, but practical backends often retain the final language-modeling head (LM-head) in BF16 or FP16. Quantizing this projection naively can strongly perturb the vocabulary-logit distribution. We present ARCHead, a packed LM-head compressor that combines a quantized low-rank core, group-wise INT4 residuals, and a low-rank correction fitted in an activation-derived metric. ARCHead stores no dense BF16 head and reduces persistent LM-head storage by 3.7-3.9x. On Qwen3-8B-Base, it uses 25.6% o…

A common task in legal Information Retrieval (IR) is to find relevant legal sources from case-law collections. While legal practice often requires pinpoint citations (pincites) to specific case paragraphs, most existing public legal IR datasets lack paragraph-level citation annotations. Yet, publicly available datasets with such information contain data leakage in the query text and exclude paragraphs that are neither citing nor cited from the corpora, creating an unrealistic and oversimplified retrieval setting, potentially leading to inflated performance. To address these limitations, we …

On-policy training has emerged as a powerful post-training paradigm for improving the reasoning capabilities of large language models, and is often enhanced by golden trajectories from stronger expert models. However, when the expert fails on harder problems, existing trajectory-guided methods lose their main source of supervision, and these failed trajectories are typically discarded as negative samples. We argue that such failures, which we call Golden Negative Trajectories, can still provide valuable reasoning signals when treated not as demonstrations to imitate, but as flawed trajector…

Chart question answering (CQA) requires multimodal large language models (MLLMs) to integrate visual comprehension with logical reasoning, yet current models struggle with accurate visual grounding and coherent reasoning chains. While extrinsic chain-of-thought prompting and visual cues significantly improve performance, current MLLMs lack intrinsic visual grounded reasoning capabilities, leading to inaccurate perception and reasoning disconnected from visual evidence. To address these limitations, we propose CURV, a curriculum learning framework that develops intrinsic visual reasoning cap…