1056 个
本仓库最初是我毕设项目的 PyTorch Lightning 模板，后来演进为包含完整项目实现本身。研究问题是地理定位如何影响遥感视觉-语言模型的训练与推理。注：仓库名称与描述可能变化。This repo originally was a PyTorch Lightning template for my thesis project. It has since evolved to contain the full project implementation itself. The research question is how geolocation influences the training and inference of vision-language models for remote sensing. Note: The repository name and description may change
無料のローカルAIインストーラ。ダブルクリック一発・初期設定なしで、チャット・RAG（文書質問）・Web検索が使えるWindows向けローカルAI環境（Ollama + Open WebUI + qwen3.5）
AI-Native 钢琴学习 RAG 助手——在线且免费（GitHub Pages -> Google Cloud Run -> Neon pgvector -> Groq）：具备引用、护栏与内置可观测性（延迟/token/成本）的摄取与查询流水线（rewrite -> 混合检索/RRF -> rerank -> LLM），数据持久化至可检索数据库。Node.js 实现，不依赖付费 API。AI-Native Piano Learning RAG Assistant - live & free (GitHub Pages -> Google Cloud Run -> Neon pgvector -> Groq): ingestion + query pipeline (rewrite -> hybrid search/RRF -> rerank -> LLM) with citations, guardrails, and built-in observability (latency/tokens/cost) persisted to a searchable DB. Node.js, no paid APIs.
生产级 LLM 网关：兼容 OpenAI 的 API，支持语义缓存，可在 Claude/OpenAI/Ollama 之间进行智能路由与成本分析。Production LLM gateway: OpenAI-compatible API, semantic caching, intelligent routing across Claude/OpenAI/Ollama, cost analytics.
使用 TurboQuant 和 MLX 在 Windows 上运行 Qwen3.5 语言模型，实现快速的本地推理。Run the Qwen3.5 language model on Windows using TurboQuant and MLX for fast local performance.
在 DABench 数据分析任务上对 DSPy RLM（Recursive Language Models）进行基准测试，使用自动评分实现基于代码的迭代评估Benchmark DSPy Recursive Language Models on DABench data analysis tasks with automated scoring for iterative code-based evaluation
将混乱的需求转化为 coding agent 一次就能做对的 prompt。一个适用于 Claude Code、Cursor 以及任何能阅读 markdown 的 agent 的 skill。Turn a messy request into a prompt your coding agent gets right the first time. A skill for Claude Code, Cursor, and any agent that reads markdown.
AI 驱动的生物医学文献综述平台：从 PubMed 检索论文，使用 FAISS 进行语义搜索，并基于 Gemini LLM 生成循证文献综述🧬 AI-powered biomedical literature review platform that retrieves PubMed papers, performs semantic search using FAISS, and generates evidence-based literature reviews using Gemini LLM.
通过查询条件超边预测，为 LLM agent 检索工具集合。Retrieve sets of tools for LLM agents using query-conditioned hyperedge prediction.
面向 CBSE、ICSE、AP SSC（1-10 年级）的 AI 备考平台 —— 基于知识点先决关系 DAG 的自适应练习、行为分析、实时竞技房间，以及减少绝大多数 AI 调用的 7 层缓存。AI-powered exam prep for CBSE, ICSE and AP SSC (Grades 1-10) — adaptive practice over a topic-prerequisite DAG, behavioural analysis, live competition rooms, and a 7-layer cache that keeps most AI calls from ever firing.
IcePaw — 本地优先的 LLM 对话工作站。多 Agent 协作、工具调用、知识库、委派任务，数据全部留在本地（Tauri + Rust + Vue）
HateMirage——可解释的伪仇恨检测与多维推理，ICON 2026 共享任务。在 HateMirage 语料库（4,530 条带标注的伪仇恨评论）上进行目标识别 + 意图与隐含意义生成。HateMirage - Explainable Faux Hate Detection and Multi-Dimensional Reasoning - Shared Task @ ICON 2026. Target identification + Intent and Implication generation over the HateMirage corpus (4,530 annotated Faux Hate comments).
借助这款轻量、原生的 Text Services Framework 输入法编辑器在 Windows 上输入孟加拉语，无需后台进程。Type Bangla on Windows with this lightweight, native Text Services Framework Input Method Editor that operates without background processes.
Experimental Qwen3.5-derived 752M LLM：基于 Qwen3.5 的实验性 752M 参数 LLM，采用 CPT + SFT 两阶段流程，面向编码、技术推理与指令跟随。Experimental Qwen3.5-derived 752M LLM: a two-phase CPT + SFT pipeline for coding, technical reasoning, and instruction following.
语言驱动、无锁的分布式工作流引擎 — 用 FFL 编写工作流，运行时负责执行、依赖解析、恢复和热部署。Language-directed, lock-free distributed workflow engine — write workflows in FFL, the runtime handles execution, dependency resolution, recovery, and live deployment.
CampusIQ 是由 AI 驱动的学术知识管理与问答系统，借助 RAG、语义搜索、OCR、图像理解以及基于 LLM 的回答生成，实现教育内容的智能检索与分析。CampusIQ is an AI-powered academic knowledge management and question-answering system that enables intelligent retrieval and analysis of educational content using RAG, semantic search, OCR, image understanding, and LLM-based response generation.
面向生态学文献综述的 AI 辅助流水线，内置对 iEcology 与保护生物学相关数据源（社交媒体追踪、博物馆标本记录、新闻档案）的支持。AI-assisted pipeline for ecological literature review with built-in support for data sources relevant to iEcology and conservation biology (social-media tracking, museum specimen records, news archives).
升级 Grand Theft Auto V（GTA V）画面表现，增强真实光照、天气效果与改进贴图，带来终极 4K 视觉体验。Upgrade Grand Theft Auto V graphics with realistic lighting, weather effects, and improved textures for a definitive 4K visual experience.
AI Kubernetes 升级智能平台——通过确定性兼容性分析 + RAG-grounded LLM 规划，实现安全集群升级（EKS/GKE/AKS/OpenShift/kubeadm）。AI Kubernetes Upgrade Intelligence Platform — deterministic compatibility analysis + RAG-grounded LLM planning for safe cluster upgrades (EKS/GKE/AKS/OpenShift/kubeadm)
由 AI 驱动的交易研究与市场情报平台，使用智能 AI agent、SQL、RAG、市场数据以及外部工具来分析研究问题并生成结构化的市场洞察。An AI-powered trading research and market intelligence platform that uses an intelligent AI agent, SQL, RAG, market data, and external tools to analyze research questions and generate structured market insights.
🤖 通过教育型对冲基金模型探索 AI 驱动的交易决策，融合多位知名金融专家的多元投资策略。🤖 Explore AI-driven trading decisions with our educational hedge fund model, featuring diverse investment strategies from renowned financial experts.
可复现研究流程，通过改造预训练 LLM tokenizer 降低尼泊尔天城文分词成本：连续/字素感知的 BPE 候选挖掘、基于边际增益的贪心词表选择、保留 ID 的 merge 拼接、基于分解的 embedding 初始化，以及 embedding-only/LoRA/CPT 适配；面向 6GB 笔记本 GPU 调优。Reproducible research pipeline that reduces Nepali Devanagari tokenization tax by retrofitting a pretrained LLM tokenizer: continued/grapheme-aware BPE candidate mining, greedy marginal-gain vocabulary selection, ID-preserving merge splicing, decomposition-based embedding init, and embedding-only/LoRA/CPT adaptation. Tuned for a 6GB laptop GPU.
Rust Graph Tracker：为 LLM 编程 Agent 提供数值与日期来源追踪，支持表达式求值与推导验证Rust Graph Tracker: numeric and date provenance tracking with expression evaluation and derivation verification for LLM coding agents
Euro Truck Simulator 2 / American Truck Simulator 的遥测与驾驶分析 MCP server，附带无头 Claude 分析 agent。只读、Family-A。MCP server for Euro Truck Simulator 2 / American Truck Simulator telemetry + driving analytics, with a headless Claude analytics agent. Read-only, Family-A.
不要再重复解决同一段代码——一个本地优先的分布式推理缓存：来自真实环境的匿名兼容性证据，加上面向编码 LLM 的已验证最小示例。Stop solving the same code twice — a local-first distributed reasoning cache: anonymous compatibility evidence from real environments plus verified minimal samples for coding LLMs.
受 MemGPT 启发、支持 OpenAI API 的长期记忆 Agent，具备记忆生命周期控制与评估能力。MemGPT-inspired long-term memory agent with OpenAI API support, memory lifecycle controls, and evaluation.
针对任意 GitHub 仓库的自然语言问答 —— 构建 Neo4j 依赖图与 Qdrant 向量索引，用自然语言即可提问结构性与行为性问题。Natural-language Q&A over any GitHub repo — builds a Neo4j dependency graph + Qdrant vector index so you can ask structural and behavioral questions in plain English
一个由 AI 驱动的科研情报平台，利用 LLM 和现代 NLP 技术，实现语义搜索、基于 RAG 的问答、文献综述生成、研究空白检测、知识图谱构建、引文管理以及科研写作辅助。An AI-powered Research Intelligence Platform that enables semantic search, RAG-based question answering, literature review generation, research gap detection, knowledge graph construction, citation management, and scientific writing assistance using LLMs and modern NLP techniques.
使用多 Agent RAG 系统与混合索引、LangGraph 编排，自动获取 arXiv 研究论文并生成文献综述。Automate arXiv research paper retrieval and literature review generation using a multi-agent RAG system with hybrid indexing and LangGraph orchestration.
通过 MegaQwen CUDA megakernel 加速 Qwen3-0.6B 推理，在 RTX 3090 上达到 531 tok/s decode，较 HuggingFace 提升 3.9×🚀 Achieve faster Qwen3-0.6B inference with the MegaQwen CUDA megakernel, delivering 531 tok/s decode on RTX 3090—3.9x faster than HuggingFace.
基于已验证决策构建的共享 intelligence 层。A shared intelligence layer built from verified decisions.
面向 Agentic Dynamics 的实验工具：衡量 AI Agent 如何行为、恢复并产出已验证的结果Experimental instrument for Agentic Dynamics: measuring how AI agents behave, recover, and produce verified outcomes
基于 PRISMA 2020 的 LLM 微调技术系统综述（2020–2025）— 84 项研究，涵盖 LoRA/QLoRA/RLHF/DPO，本科毕业论文（PUCE）。PRISMA 2020 systematic review of LLM fine-tuning techniques (2020-2025) - 84 studies, LoRA/QLoRA/RLHF/DPO, undergraduate thesis (PUCE).
混合 RAG 系统，具备 reranking、基于引用的 grounded citations，以及可选 provider-backed 执行的可确定离线评估分级。A hybrid RAG system with reranking, grounded citations, and deterministic offline evaluation tiers with optional provider-backed execution.