Moderate Embeddings Question 76 of 223

Why do teams L2-normalize embeddings?

GenAI / LLM · Speak this in 60–90 seconds · Faridabad & Delhi NCR

PICTURE THIS: GIT FLOW

Working folderYour files
Staginggit add
Local repogit commit
Remotegit push

Simple meaning

After unit normalization, inner product equals cosine similarity, which simplifies indexes and scores.

1

WHY — Tokens instead of words?

LLMs use tokens (not full words) because it helps them:

They are checking judgment

on Embeddings.

A good answer names

the situation, the default choice, and one exception - that reads as experience.

Stable token IDs

Each piece maps to a number the network can learn.

Fits the model

Fixed pieces are what transformers expect as input.

2

STEPS — What happens with tokens?

Before the model can read a sentence, it goes through these steps:

  1. 1
    After unit normalization, inner

    product equals cosine similarity, which simplifies indexes and scores.

  2. 2
    It also reduces length

    bias where longer chunks look closer just because their vectors are larger.

  3. 3
    Confirm the embedder was

    trained with the same convention.

  4. 4
    Context mix

    Attention looks at nearby tokens together.

  5. 5
    Next token

    The model scores what should come next.

  6. 6
    Decode

    IDs turn back into readable text.

3

EXAMPLE — See it in action

Let's see how a real sentence is tokenized (tokens may vary by model):

Input text
“It also reduces length bias where longer chunks look closer just because their v”
Tokenized output
Italsoreduceslengthbiaswhere
Token IDs (example)
2987408337471632900

Note: Actual tokens and IDs depend on the tokenizer (e.g., GPT, Llama, etc.).

Key takeaway

After unit normalization, inner product equals cosine similarity, which simplifies indexes and scores. It also reduces length bias where longer chunks look closer just because their vectors are larger.

Chat with us