LLM Interview Handbook

What LLM Interview Handbook does

The README, titled 大模型面试宝典, organises its material by exam topic rather than by article or company. Modules include RAG (160 topics), Agent (100), reinforcement-learning post-training (120), Hermes Agent (100), Loop Engineering (100), GraphRAG (100) and object detection (100), plus more than 30 classic topics. Each question is written the same way: a one-sentence answer, an explanation of why, and the follow-up questions an interviewer is likely to ask.

The Hermes Agent module has 22 questions with full write-ups and answer points so far. A separate Harness Engineering section covers the idea of Agent = Model + Harness, routing, memory and permissions, and self-evolution topics such as automatic skill generation and a Nudge Engine. Readers can open issues to suggest new questions or send pull requests to fix answers. The repository began as a collection of NLP paper-study notes, and the README is written in Chinese.

Key features

• More than 800 high-frequency interview topics across LLM, RAG, Agent and RL
• 100-topic Hermes Agent module, with 22 detailed write-ups so far
• One-line answer, deeper breakdown and likely follow-ups for each question
• Difficulty ratings from one to three stars
• Extra modules for Loop Engineering, GraphRAG and object detection
• Questions via issues and corrections via pull requests

When to use it

• Prepare for an AI Agent or LLM algorithm interview in Chinese
• Revise Hermes Agent concepts through interview-style questions
• Pick graded questions as an interviewer

Who it is for: Chinese-speaking candidates preparing for LLM and AI Agent interviews, and interviewers who need graded questions.

How it fits with Hermes Agent

A study resource that includes a dedicated 100-topic Hermes Agent module among several other AI topics; it is not a Hermes-specific project.

Note: The material is in Chinese, and only 22 of the 100 Hermes Agent questions have detailed write-ups so far.

FAQ