2026年9月10日 技术热点总结

项目简介：Use Astra as orchestrator and Luna for subagents in Codex

技术栈：Python

核心介绍：A configurable Codex setup where GPT-6 Astra is the root/orchestrator and reviewer, while GPT-5.6 Luna is the default and pinned model for execution subagents.

项目数据：⭐ 717 Stars，🍴 55 Forks

项目简介：Here is a dlssg for RTX30 Series GPU

核心介绍：简体中文 | English Windows x64 / D3D12。运行文件为 version.dll 和 dlssg_sm86.ini。 自有 C++ 包装层、SM75/SM86 PTX/Cubin、310.1 模型和推理图都在一个 DLL 内。运行时不解压、加载或内存映射原厂 nvngx_dlssg.dll；仍使用系统 NVIDIA NGX/NVAPI/CUDA 驱动接口，无需 CUDA Toolkit。 以下为 0.2.3 在 RTX 3080 Ti / 驱动 591.86 / PTX 下的参考预算，覆盖 SM86 默认精确、SM86 可选近似及 SM75 路由，共 27 组补测。按最终输出分辨率选行；例如 4K 输出 + DLSS 性能档仍使用 4K 这一行。

项目数据：⭐ 657 Stars，🍴 39 Forks

项目简介：手绘风格编号画廊与双语提示词 Skill

技术栈：HTML

核心介绍：不会描述画风，也能快速做出有辨识度的图片。 这里整理了 001–261 种手绘风格。先选一个编号，再告诉 Skill 你想表现的主题，就能得到带有风格名称的中文和英文提示词，直接交给支持生图的 AI 使用。 很多人只能说“可爱一点”“文艺一点”，每次生成结果却不一样；同一个主题换一次 AI，画风也会漂移；面对一整页参考图，又很难准确说出自己喜欢哪一种。 这个风格库把“凭感觉选图”变成“记住一个编号”。你不需要背诵画风名称，也不需要自己拼提示词。 1. 打开编号画廊，浏览风格图片。 2. 记下喜欢的编号，例如 041。 3. 输入“编号 + 主题”，例如：041号风格，主题：秋天的第一杯奶茶。 4. 得到带风格名称的中文和英文提示词。 5. 复制到你正在使用的生图 AI 中。 Skill 默认只负责把想法变成提示词；你明确要求“生图”时，会先判断当前生图模型能否仅凭风格名称激活画风。能激活就不传参考图，不能确认或不能激活就传对应编号的单图，避免模型把参考图中的主体和构图一起锁死。

项目数据：⭐ 626 Stars，🍴 86 Forks

We introduce the Graph Machine (GM), an architecture that maintains an O(n)-sized state and accesses it through sparse, dynamic routing. Unlike methods with fixed-size states or sparse but static routing, GM preserves O(n) complexity in its sparse layers without restricting the potentially accessible state size to O(1). Instead, GM uses edges – pointer-like objects updated differentiably by a referral mechanism resembling pointer chasing. We replace 75% of the dense Transformer layers in Qwen3-0.6B with GM sparse layers and pretrain from scratch on 15.7B tokens. With only 2 of 4,096 tokens …

Modern LLM-based agents operate through a harness of tools, reusable skills, and specialist agents that shapes what they observe and what they can do. In practice, this harness continually evolves as new capabilities are added. We introduce EVOHARNESSBENCH, a benchmark for evaluating agents under controlled harness evolution across three axes (tools, skills, and agents). Unlike existing continual-learning benchmarks for agents, which typically place non-stationarity (i.e., what changes over time) in the task stream while keeping the harness fixed, EVOHARNESSBENCH places non-stationarity in …

We present 'RenderFormer-V2', a unified learned transformer-based neural rendering model, complementary to modern physics-based rendering systems, that can handle diverse light-transport effects such as caustics, volumetric scattering, environment lighting, textured and displaced surfaces and out-of-distribution materials without per-scene training or specialized code. RenderFormer-V2 models global light transport as a sequence-to-sequence transformation. Following its predecessor, RenderFormer-V2 also employs a two stage process: a view-independent stage that resolves intra-scene primitive…

Multimodal Large Language Models (MLLMs) perform strongly on general visual understanding tasks such as visual question answering, yet they often struggle with a basic comparative skill: identifying what has changed between two similar images. We introduce VDiff-Bench, a challenging multiple-choice benchmark for fine-grained Image Difference Identification. VDiff-Bench contains 1,756 four-way questions over image pairs and covers 10 change categories: position, motion, regional image color, overall image color, appearance/disappearance, noise/resolution, texture, substitution/size, OCR/text…

The digitization of healthcare has generated vast, longitudinal, and multimodal patient records over a lifetime, yet fully exploiting these data to represent and predict patient state trajectories remains a critical challenge. Current AI models often struggle to capture the complex, irregular temporal dynamics and inherent stochasticity of real-world multimodal patient data. Existing AI approaches for modeling longitudinal patient records are predominantly discriminative, limited to a few modalities, constrained by closed categorical vocabularies, treating time as a monotonic inductive bias…

We study a governed approach to enterprise analytics: a language model interprets the question, while deterministic policy selects and runs a pre-approved analytical program that returns both results and evidence. We show that this restriction can remain expressive within a defined analytical class, using relational operations plus aggregation, comparison, windows, ranking, and similarity. Fixed meaning, policy, data, and execution rules also make results replayable. Across 440 runs, three 8B models generated SQL and selected tools at runtime, while Qwen3-8B interpreted intent only and poli…