The main purpose is to list all the resources I’ve collected about Machine Learning (ML), Artificial Intelligence (AI), and Large Language Models (LLM).

ML Basics

For a intro to ML, I recommend the following resources:

Courses:

• Google Machine Learning Crash Course
• Stanford CS229: Machine Learning by Andrew Ng

Videos:

• 3Blue1Brown’s intro to neural network
  • The youtube channel also has a lot of good videos about math (linear algebra, calculus, etc.), physics, and a lot of science topics.
• Andrew Karpathy’s Neural Networks: Zero to Hero

There are many more excellent introductory materials.

LLM

The two best courses for LLM are:

• Stanford CS224n (winter 2025)
  • A very good course for a lot of intuitions for LLM with Youtube videos
• Stanford CS336 (spring 2025)
  • The most hard-core course for LLM with Youtube videos

Videos:

• Chinese: GPT 1-3 paper read by Mu Li
  • A famous researcher and youtuber, who has good videos about Transformer, GPT1-4, InstructGPT, CLIP, GAN, Whisper, etc.
• Google DeepMind: How to Scale Your Model

Training Infra

• Introduction to LLM Parallelism
  • A good introduction to the parallelism in LLM training from myself.
• The Ultra-Scale Playbook: Training LLMs on GPU Clusters
  • Hugging Face’s playbook for training LLMs on GPU clusters. Very comprehensive and detailed.

Reinforcement Learning

• RL book from Sutton and Barto (2020)
• Youtube videos from EZ.Encoder
  • Excellent videos about Deepseek, RL, etc.

Interpretability

• A Mathematical Framework for Transformer Circuits (Dec 2021)
  • Anthropic’s paper conceptualizing the operation of transformers in a new but mathematically equivalent way and making sense of these small models and gain significant understanding of how they operate internally.
• In-Context Learning and Induction Heads (Mar 2022)
  • Anthropic’s paper arguing that induction heads may be the mechanistic source of general in-context learning in transformer models of any size.
• Towards Monosemanticity: Decomposing Language Models with Dictionary Learning (Oct 2023)
  • Anthropic’s paper using a sparse autoencoder to extract a large number of interpretable features from a one-layer transformer.
• Sparse Autoencoders Find Highly Interpretable Model Directions (Oct 2023)
  • Using sparse autoencoders to find meaningful directions within a model’s activations.
• Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet (May 2024)
  • Applying and scaling these ideas to larger, more capable models like Claude 3 Sonnet.
• Scaling and evaluating sparse autoencoders (Jun 2024)
  • Exploring the practical aspects and effectiveness of sparse autoencoders at scale.
• Transcoders find interpretable LLM feature circuits (Jun 2024)
  • Focusing on finding interpretable circuits of features within LLMs.
• On the Biology of a Large Language Model (Mar 2025)
  • Anthropic’s most recent paper for LLM Interpretability.

Just like you can infinite scroll on TikTok, you can infinite scroll on the papers.

Agent

• UCB CS294/194-196 Large Language Model Agents
  • A good course from UC Berkeley with Youtube videos, which invited a lot of frontier researchers to give lectures.
  • List of topics
    • Inference-Time Techniques & Reasoning (CoT, ReAct, RAG, Planning, etc.)
    • Coding Agents
    • Multimodal Autonomous AI Agents
    • AlphaProof, Science Discovery
    • Reinforcement Learning
    • Safety & Vulnerability
    • etc.
  • I’m planning to write a summary for this course.

###

All About Transformer Inference

Recent LLM Papers (that I read and liked)

Need to mention that a lot of the courses and resources already include a lot of good papers.

• Speculative deconding papers
  • Fast Inference from Transformers via Speculative Decoding (Nov 2022)
  • Accelerating Large Language Model Decoding with Speculative Sampling (Feb 2023)
  • DistillSpec: Improving Speculative Decoding via Knowledge Distillation (Oct 2023)
  • Decoding Speculative Decoding (Feb 2024)

Vision

Intro:

• Image Generation 101: an Introduction A good introduction to the image generation from myself.

Courses:

• Stanfodd CS231n
  • There seemse to be videos only from 2017

Papers

• The intro listed a lot of important papers.

• Large Language Models are Zero-Shot Reasoners (May 2022)

  • The famours “Let’s think step by step” paper.

News & Blogs

• Lil’Log
  • Lilian Weng’s blog, ex VP of Research at OpenAI.
• Chinese: I usually listen to 大飞
• Chinese: A good one is aidaily.win

• AI Weekly

• The bitter lessons Mar. 2019, Richard Sutton.
  • The biggest lesson that can be read from 70 years of AI research is that general methods that leverage computation are ultimately the most effective, and by a large margin.
• The Second Half, Apr. 2025, Shunyu Yao.

• Welcome to the Era of Experience, Apr. 2025, David Silver & Richard Sutton, DeepMind.

• Conciousness, strongly recommend to listen to Jeffrey Hinton’s talk in 2024.
  • Will Digital Intelligence Replace Biological Intelligence Nov 2024.
• CBMM10 Panel: Research on Intelligence in the Age of AI Panel discussion of Jeffrey Hinton, Demis Hassabis, Illya Sutskever.

Terms

• Attention
• Chain of Thought

• Flash Attention

• ReAct
• Transformer

Old

• SIFT features: Scale-Invariant Feature Transform, old Visision method, outdated.

People

Jeffrey Hinton, Demis Hassabis, Ilya Sutskever,

Yoshua Bengio, Ian Goodfellow, Yann LeCun, Meta Richard Sutton

Sam Altman, Dario Amodei, Fei-Fei Li, Andrew Ng, David Silver,

Andrew Karpathy Jared Kaplan, Anthropic codouner and CSO.