仓库/Skill 库

找到在你的硬件上真正能跑且性能最优的本地 LLM。排名基于真实且时新的基准测试，而非参数量。一条命令，即刻运行。Find the local LLM that actually runs and performs best on your hardware. Ranked by real, recency-aware benchmarks, not parameter count. One command, run it instantly.

大型多模态模型 LMM 的开源评估工具包，支持 220+ LMM 与 80+ 基准测试。Open-source evaluation toolkit of large multi-modality models (LMMs), support 220+ LMMs, 80+ benchmarks

🦀️ CRAB: 面向多模态语言模型 Agent 的跨环境 Agent 基准。https://crab.camel-ai.org/🦀️ CRAB: Cross-environment Agent Benchmark for Multimodal Language Model Agents. https://crab.camel-ai.org/

[COLM 2025] 论文 Know Me, Respond to Me: Benchmarking LLMs for Dynamic User Profiling and Personalized Responses at Scale——大语言模型在动态用户画像与大规模个性化响应任务上的基准测试。[COLM 2025] Know Me, Respond to Me: Benchmarking LLMs for Dynamic User Profiling and Personalized Responses at Scale

一个 LLM 相关论文、学位论文、工具、数据集、课程与基准的合集。A collection of LLM related papers, thesis, tools, datasets, courses, benchmarks

面向长时对话记忆层的综合基准测试框架A Comprehensive Benchmarking Framework for Long-Term Conversational Memory Layers

实时更新的向量数据库项目、集成和基准评测全景图——每……刷新。Live-updating landscape of vector database projects, integrations, and benchmarks — refreshed every

AI agent memory 与基础设施全景——912 个系统 × 68 列的对比目录，覆盖记忆层、agent 框架、运行时、vector store、知识图谱、MCP server、benchmark。支持按类型化边、谱系、引用进行检索。AI agent memory & infrastructure landscape — comparative catalog of 912 systems × 68 columns covering memory layers, agent frameworks, runtimes, vector stores, knowledge graphs, MCP servers, benchmarks. Searchable with typed edges, lineages, citations.

🔍 在 Weaviate 混合检索中评测 embedding 模型，基于自有数据或 MTEB 数据集评估 MRR@K、Hit@K、延迟与内存占用🔍 Benchmark embedding models in hybrid search with Weaviate. Evaluate MRR@K, Hit@K, latency, and memory using your data or MTEB datasets.

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

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).

追踪并比较每一个严肃的 OpenClaw 替代方案——基于实测仓库数据与 AI 撰写的决策支持，且刻意保持独立。Track and compare every serious OpenClaw alternative — measured repo data, AI-written decision support, kept apart on purpose.

🚀 使用强化学习优化半精度通用矩阵乘法（HGEMM）CUDA kernel，性能超越 cuBLAS 及其他基准。🚀 Optimize Half-precision General Matrix Multiply (HGEMM) CUDA kernels using reinforcement learning, surpassing cuBLAS and other benchmarks with superior performance.

RAMR —— 检索增强记忆可靠性：面向 Agentic-RAG / 记忆系统的抗污染合成基准（附方法与发现）。RAMR — Retrieval-Augmented Memory Reliability: a contamination-resistant synthetic benchmark for agentic-RAG / memory systems (findings + method)

本地 LLM 基准测试、RAG、语音交互与 AI 助手Local LLM benchmarking, RAG, voice interaction and AI assistant

🎨 利用 AI 实时生成高保真 UI 设计，对比多版本方案，并跨多个模型导出可直接使用的代码。🎨 Generate high-fidelity UI designs in real-time with AI, compare variations, and export ready-to-use code across multiple models.

基于人类基准的 LLM 生成 Infrastructure-as-Code 安全基准测试。100 场景 × 12 模型配置 = 1,196 个工件，由 Checkov/Trivy/KICS 扫描，并与 634 个人工编写模板对比。所有模型的漏洞密度均为人工的 3.2–3.9 倍。arXiv:2608.28021Human-anchored security benchmark for LLM-generated Infrastructure-as-Code. 100 scenarios × 12 model configs = 1,196 artifacts scanned by Checkov/Trivy/KICS, compared against 634 human-written templates. Every model: 3.2–3.9× human vulnerability density. arXiv:2608.28021

面向文档溯源 QA Agent 的生产级 Harness：检索、起草、自评、重写、升级循环，配套 LLM-as-judge 评测套件、分级权限工具、对破坏性操作的人工审批，以及一键容器化部署。A production-grade harness for a document-grounded QA agent: a retrieve, draft, self-score, re-draft, escalate loop with an LLM-as-judge eval suite, permission-tiered tools, human-in-the-loop approval for destructive actions, and a one-command container deploy.

面向 AI coding agents 的 token 预算 context pack，基于编译器解析的 Kotlin 结构（Analysis API/PSI）构建，并附带衡量其是否优于 chunk RAG 的 benchmark。Token-budgeted context packs for AI coding agents, built from compiler-resolved Kotlin structure (Analysis API/PSI) — with the benchmark that measures whether it beats chunk RAG