🌱 ARIS in AI Offereasyaioffer.github.io ⭐ Star

AI 秋招面试 Cheat Sheet,一个链接看全部AI-interview cheat sheets — everything from one link

每篇:公式推导 + 可运行代码;绝大多数附 25 道高频题(L1 必会 / L2 进阶 / L3 顶级 lab),题数见卡片。中英双语,单文件 HTML,手机上直接读。Each sheet: derivations + runnable code; nearly all carry 25 interview questions (L1 / L2 / L3) — the count is on the card. Bilingual, single-file HTML, reads fine on a phone.

35 篇sheets8 个方向categories中 / EN 双语bilingual短地址short address easyaioffer.github.io跨模型审查后发布cross-model reviewed
已读Read 0/35

🧠 基础Foundations 7

🎯 后训练与推理Post-Training & Reasoning 4

🏛️ 架构与系统Architecture & Systems 8

长上下文Long Context

RoPE · YaRN · NTK · MLA · StreamingLLM · 位置外推RoPE · YaRN · NTK · MLA · StreamingLLM · position extrapolation

25 题QEN中MDMD
量化Quantization

GPTQ · AWQ · SmoothQuant · FlatQuant · FP8/NVFP4 · QAT: LSQ·ParetoQ·Gemma-QAT · FP4 训练GPTQ · AWQ · SmoothQuant · FlatQuant · FP8/NVFP4 · QAT: LSQ·ParetoQ·Gemma-QAT · FP4 training

25 题QEN中MDMD

🌊 生成模型理论Generative Theory 3

🎨 生成系统Generation Systems 6

视频生成Video Generation

Sora · Hunyuan-Video · Kling · Wan · Movie Gen · 时空 token 化Sora · Hunyuan-Video · Kling · Wan · Movie Gen · spatio-temporal tokenisation

25 题QEN中MDMD
3D 生成3D Generation

NeRF · Instant-NGP · 3DGS · SDS · DUSt3R→VGGT · 原生 mesh · TRELLIS/Hunyuan3D 2.x · sim-readyNeRF · Instant-NGP · 3DGS · SDS · DUSt3R→VGGT · native mesh · TRELLIS/Hunyuan3D 2.x · sim-ready

25 题QEN中MDMD

👁️ 多模态Multimodal 1

VLM / 多模态VLM / Multimodal

CLIP · LLaVA · Qwen-VL · DeepSeek-VL · 视觉 token · 对齐训练三阶段CLIP · LLaVA · Qwen-VL · DeepSeek-VL · visual tokens · three-stage alignment

25 题QEN中MDMD

🤖 AgentAgents 5

Agentic RLAgentic RL

AgentTuning · ToolRL · RAGEN · WebRL · SWE-RL · 多轮 GRPOAgentTuning · ToolRL · RAGEN · WebRL · SWE-RL · multi-turn GRPO

25 题QEN中MDMD

🦾 具身与世界模型Embodied & World Models 2

第一方的 World Models 篇在前;具身题库由社区作者维护、外站托管,有帮助的话去源仓库点个 ⭐ 谢谢作者。The World Models sheet is first-party; the embodied Q&A bank is community-maintained and hosted elsewhere — star the source repo to thank the author.

📝 长文 BlogLong-form Blogs 3

三篇中文长文:世界模型架构、连续扩散语言模型综述、扩散与表征学习。Three Chinese long reads: a world-model architecture, a survey of continuous diffusion LMs, diffusion meets representation learning.

没有匹配的教程。换个关键词试试,或者点「全部」。Nothing matches. Try another keyword, or tap “All”.