AI Systems · advanced
Self-Attention
Q/K/V, scaled dot-product, causal masking.
Mental model
Self-attention lets every token look at every other token and pull in what is relevant. Each token emits a query, a key, and a value; the query-key dot products (scaled, softmaxed) become weights over the values. Causal masking blocks a token from seeing the future.
How to study Self-Attention
Begin by restating the mental model in your own words, then connect it to a concrete system you have built or operated. Name the mechanism, the constraint it addresses, and the trade-off it introduces. Use Stanford CS336 — L6 Kernels & Triton (FlashAttention2), CS224n — Self-Attention & Transformers (Stanford), The Illustrated Transformer to check details, but close the source before writing your explanation. Retrieval is the learning step; rereading is only preparation.
Next, compare Self-Attention with Multi-Head Attention. Ask what changes in correctness, latency, resource use, operability, and failure recovery. Complete Scaled dot-product attention weights and preserve the command, input, output, and one failed attempt as evidence. Finish by explaining the idea without jargon to someone who has not studied the track.
Proof of understanding
- Explain the mechanism from first principles and identify the state it reads or changes.
- Give one situation where the concept is the right choice and one where it is not.
- Predict a realistic failure mode before running the drill, then compare the prediction with evidence.
- Connect the result to a roadmap or build artifact instead of treating the concept as isolated trivia.
Common mistakes
- Forgetting the 1/sqrt(d_k) scaling, which destabilizes the softmax
- Omitting the causal mask in a decoder, leaking future tokens
- Confusing the query/key/value roles
Learn from primary sources
Practice and explain it back
Scaled dot-product attention weights
Q=K=[[1,0],[0,1]], V=[[1,2],[3,4]], d_k=2. Compute softmax(QKᵀ/√d_k)·V for one query row.
Expected evidence: Uniform weights 0.5/0.5 → output [2,3].
Open the interactive drill →Review prompts
- Why divide the query-key dot product by sqrt(d_k)?
Build evidence
Synthesize: AI Models & Training
Move from transformer foundations through pre-training, fine-tuning, post-training, compression, and evaluation. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks