What does the dot product q . k actually measure inside an attention layer?

Attempt it

It measures unnormalized directional alignment between q and k. Large positive when the two vectors point the same way, near zero when they are roughly orthogonal, negative when they point in opposite directions. Magnitude also grows with the lengths of q and k. After dividing by sqrt(d_k) to stabilize variance and passing through softmax, that raw score becomes the attention weight the query assigns to that key. So q . k is the affinity signal; scaling and softmax just turn it into a probability.

Your answer

Quick reference

TL;DR

q dot k is the unnormalized affinity score between a query and a key, combining direction (cosine of the angle) with vector magnitude.

Memory aid

Sign in for the mnemonic that makes this stick. See Pro pricing.

Easy to grasp

Picture two people at a job fair pointing flashlights at each other. The query is your flashlight, the key is theirs. When you both shine straight at each other, the beams overlap a lot, strong match. When you face perpendicular directions, the beams barely cross, weak match.When you face away, the beams point apart, negative match. The dot product is just a number that says how much the two beams overlap. The model later turns that overlap into a probability and uses it to decide how much of the key's payload to pay attention to.

Key concepts

Dot product Cosine similarity Softmax

Detailed answer & concept explanation

~5 min read

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example. Click to expand.

Everything you need to truly understand this topic: intuition, mechanics, step by step explanation, code, formulas, and worked example.

Sign in for 5 free unlocks/day or go Pro for unlimited. See Pro pricing.

Quick scan · 2–4 min

Everything important, quickly.

Define the dot product geometrically, name the three alignment regimes, explain why magnitude also matters, motivate the sqrt(d_k) scaling via variance, and connect to softmax turning scores into weights.

Real-world examples

Real products, models, and research that use this idea.

Every transformer-based LLM in production (GPT-5.5, Claude Opus 4.7, Llama 4 Maverick, Gemini 3.1 Pro, Qwen 3.5) computes attention via batched `Q K^T` dot products.
FlashAttention-2 and FlashAttention-3 fuse the `Q K^T` dot product with the softmax and the value matmul to avoid materializing the full score matrix in HBM.
Vector database lookups (Pinecone, Weaviate, pgvector) often default to dot-product or inner-product scoring for the same composability and hardware-acceleration reasons.

Interviewer follow-ups

What an interviewer would ask next. Try answering before peeking at the approach.

QWhy exactly `sqrt(d_k)` for the scaling, not `d_k` or some other function?

Sign in or go Pro to see the approach for this follow-up.

QIf you replaced the dot product with cosine similarity, what would change at training and serving time?

Sign in or go Pro to see the approach for this follow-up.

Don't say this

Red flags and common mistakes that signal junior thinking. Click to expand.

Most common mistake

Conflating the raw dot product with cosine similarity. They share direction information, but the dot product also scales with vector magnitudes, that magnitude effect matters.

Sign in or go Pro to see all red flags and common mistakes. See Pro pricing.

60-second cheat sheet

The night-before-the-interview bullets. Scan these on the way to the call.

Sign in for 5 unlocks/day or go Pro for unlimited revision sheets. See Pro pricing.

References

Primary sources. Skim if you want the original framing.

Vaswani et al., Attention Is All You Need