LLM-Algorithm-Intern-Guide

Why we included this project

Preparing for a large-model algorithm internship means facing a wall of scattered interview questions, and this notebook collects the ones that actually show up. It is one author's study log, built from questions asked in real interviews and gathered from Chinese tech communities, then organized into chapters that track the usual interview syllabus: Transformer internals and attention variants, RoPE and KV-cache math, LoRA and QLoRA fine-tuning, RLHF with PPO/DPO/GRPO, and close reads of the DeepSeek-V3/R1 and Qwen technical reports. The curation and the depth of explanation are the point, not runnable tooling, so treat it as a structured revision map and a source of likely follow-up questions. Since the full notes live in an external Feishu document, the repo works best as a navigable table of contents and a checklist of topics to master before the interview.