LLMs Research Lab

LLM Flashcards.

An illustrated deck for revising large language model concepts. Hand-drawn diagrams and plain-English explanations, designed for engineers prepping interviews and students revising for exams.

LLM Deck preview — 300+ visual flashcards for LLMs. Three sample cards shown: What Is a Transformer, Encoder-Only Models, and Query/Key/Value Vectors.

What’s in the deck

  1. I.Transformer architecture. Attention, FFN, layer norm, residuals.
  2. II.Attention variants. MHA, GQA, MQA, sparse, sliding-window.
  3. III.Tokenization & embeddings. BPE, positional encoding, RoPE.
  4. IV.Pretraining. Objectives, scaling laws, data mixtures.
  5. V.Fine-tuning & alignment. SFT, RLHF, DPO.
  6. VI.Retrieval & tools. RAG, agents, function calling.
  7. VII.Inference. KV cache, quantization, speculative decoding.
  8. VIII.Evaluation. Benchmarks, reasoning, contamination.

Get the deck on Gumroad →