论文复现

500 行 TypeScript 写成的深度研究 agent 参考实现，用来替换你自己拼的搜索加抓取加 LLM 循环

盲脸修复的参考实现，用来替换 GFPGAN/GPEN 这类老一代人脸复原模型

把「读论文/财报 → 提假设 → 写代码跑实验」这条研发链路做成可循环的智能体框架，顶替人工反复调参和复现

在 CI 或脚本里批量拉取 App Store 安装包，替代手动用 iTunes 抓包

把论文里的注意力、重参数、MLP 模块抄成能直接 import 的 PyTorch 层，省掉自己对着公式重写一遍

LightGlue: Local Feature Matching at Light Speed (ICCV 2023)

Scenic: A Jax Library for Computer Vision Research and Beyond

[PAMI'23] TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving; [CVPR'21] Multi-Modal Fusion Transformer for End-to-End Autonomous Driving

SALMONN family: A suite of advanced multi-modal LLMs

Pytorch implementation of Transfusion, "Predict the Next Token and Diffuse Images with One Multi-Modal Model", from MetaAI

SerpScrap retrieves structured search results for SEO, research, and automation. It is available as a Python package, a CLI, a Docker application, and an MCP-compatible server. Results are JSON-compatible and preserve the source engine, rank, URL, title, snippets, result type, and typed failures.