2026年8月4日 技术热点总结

项目简介：A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.

技术栈：C、avx2、c99、cpu-inference、deep-learning、from-scratch、inference-engine、kimi-k3、linear-attention

核心介绍：kimi-k3-in-c A 2.78-trillion-parameter model. One CPU. 8 GB of RAM. Kimi K3 inference in portable C99. No BLAS. No framework. No GPU. 2.78T parameters 1.56 TB checkpoint on disk 8.24 GB peak RSS, measured 176 KB the whole engine 0 $ ./bin/k3 ~/k3model –trunk ~/k3trunk –preset laptop + “The Eiffel 8 tokens in 261.5 s, 32.6…

项目简介：竹知了 —— 一转就哇哇叫的传统玩具，Web 模拟版。零依赖单文件，真实录音采样，移动端优先。

技术栈：HTML

核心介绍：一转就”哇哇”叫的传统玩具，Web 模拟版。零依赖单文件，手机优先。 > 甩两下要是听见了小时候那声”哇——哇——”，顺手点个 ⭐ Star 吧。 直接用浏览器打开 index.html 即可（无需构建、无需联网）。 在项目目录起个静态服务，手机连同一 Wi-Fi 访问 http://:8123： python3 -m http.server 8123 竹筒一端蒙竹膜，膜心系一根涂了松香的线，线的另一头拴在小竹签上。 甩起来转圈时，线在松香上”黏–滑”交替摩擦，脉冲沿线传到竹膜， 膜与筒腔共鸣放大——就是那声”哇——哇——”。 主音源是真实竹知了的录音采样：从实拍视频里截取 1.72 秒（恰好 4 个”哇”周期、 包络边界自动搜索对齐），尾部 50ms 等功率交叉淡化烘进头部做成无缝循环， 以 AAC 内嵌在 HTML 里保持单文件。回放速率随甩动转速变化 （录音里的甩速约 2.33 圈/秒，甩得越快叫得越急越高），再叠每圈相位的音高微摆。 采样解码失败时回退到纯合成链： 竹筒是绳系质点（重力 + 只拉不推的弹性绳 + 空气阻力，1/240s 定步长积分）。

项目数据：⭐ 751 S…

项目简介：An interactive 3D human anatomy explorer built using threejs with GPT 5.6 Sol

技术栈：TypeScript

核心介绍：A clean full-stack starter running on vinext, with optional Cloudflare D1 and

项目数据：⭐ 741 Stars，🍴 210 Forks

项目简介：An AI-native office suite for macOS and Windows: word processor, spreadsheet, presentations, and PDF.

技术栈：TypeScript、ai、docx、electron、office-suite、pdf、pptx、xlsx

核心介绍：An AI-native office suite for macOS and Windows: word processor, spreadsheet, presentations, and PDF — five Electron apps sharing one engine layer, built around AI editing as a first-class workflow rather than a bolted-on chat box. Signed installers built from main: Other versions are on the Releases page.

项目数据：⭐ 620 Stars，🍴 81 Forks

Post-training alignment is often shallow, eroding under fine-tuning. Whether midtraining interventions, cleanly isolated from post-training, can produce durable alignment remains untested. We test this via constitutional midtraining: inserting principled, values-based content into midtraining against a replay-only control at 120B scale. Our 394M-token constitutional corpus, built from Anthropic's Constitution, uses a 2×2 factorial design (curriculum ordering x deliberative reasoning) to produce four constitutionally midtrained conditions plus a control, evaluated on self-generated and estab…

The web is increasingly accessed by AI agents rather than humans. Every agent needs knowledge, especially in the life-sciences, where agentic pipelines are growing fast. Access to the literature is a crucial part of that need, and resources such as Europe PMC, with over 40M indexed records, are widely used to meet it. Yet these resources were not built for AI agents: they take keywords and complex syntax and return whole papers, so every agent must learn the syntax, issue several searches, and read full papers to find the evidence it needs. We introduce EMBL AI Librarian, a knowledge layer …

Emotional dialogue research includes two influential strategy traditions. Empathetic dialogue prioritizes understanding a speaker's emotional experience. Emotional support conversation selects and sequences support for the seeker's current needs. Sustained use introduces a further goal. Effective support should sustain users' capacities for emotion regulation, coping, self-endorsed decisions, and social connection across the interaction lifecycle. We propose capability-sustaining emotional dialogue (CSED) as a longitudinal research paradigm that aligns supportive strategy with this goal and…

Reinforcement learning with verifiable rewards (RLVR) broadcasts a single response-level reward to every token, while on-policy distillation (OPD) scores each token against a stronger teacher for a dense advantage but caps performance at teacher quality and discourages exploration beyond it. Their complementarity makes combining RLVR and OPD promising, but we find that fusing the two advantages with a fixed coefficient triggers entropy collapse from two miscalibrations: a magnitude mismatch, where token-level OPD advantages can spike far beyond the bounded RLVR advantage and erase its signa…

Despite the impressive visuomotor capabilities enabled by Vision-Language-Action (VLA) models, their performance often degrades on challenging and out-of-domain tasks. Recent test-time steering and scaling methods improve performance without extensive data collection and retraining, but action samples often remain concentrated around similar behaviors and therefore inherit correlated failure modes. Moreover, existing methods apply the same intervention strategy at every timestep, regardless of whether the base policy is already likely to succeed. To address these limitations, we introduce R…

Reinforcement learning with verifiable rewards (RLVR) is central to improving long-CoT reasoning in large language models. Critic-free methods such as GRPO convert response-level rewards into advantages and uniformly broadcast them across tokens, overlooking their unequal contributions to the final outcome. On-policy self-distillation (OPSD) instead provides dense distributional supervision by minimizing the forward KL divergence between an unprivileged policy and a privileged self-teacher, implicitly assuming that the resulting likelihood shifts encode reliable answer-aligned information. …