High Evals Question 180 of 223

How do you deal with high variance in LLM eval scores?

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

PICTURE THIS: DATA SPLIT

Train 70%Val 15%Test 15%

Fit on train, tune on val, report on test once.

Simple meaning

Fix temperature and seeds where possible, average multiple samples, and use paired tests on the same items.

WHY — Tokens instead of words?

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

This is a process

question about Evals.

Panels listen for order,

trade-offs, and what you would actually do on a GenAI / LLM project - not buzzwords.

Stable token IDs

Each piece maps to a number the network can learn.

Fits the model

Fixed pieces are what transformers expect as input.

STEPS — What happens step by step?

Before you speak the answer, walk the interviewer through these steps:

1
Fix temperature and seeds
where possible, average multiple samples, and use paired tests on the same items.
2
Report confidence intervals, not
a single percentage.
3
Small golden sets can
flip 'better' just from judge noise.
4
Give an example
One tiny concrete case you can say aloud.
5
Common mistake
What juniors usually get wrong.
6
Close
When you pick this over the alternative.

EXAMPLE — See it in action

Here's a short line you can speak, broken into clear beats:

Say this line

“Report confidence intervals, not a single percentage.”

Break into beats

Reportconfidenceintervalsnotasingle

Speaking order

2987408337471632900

Note: Adapt this scaffold to your own project — keep it under 60–90 seconds.

Key takeaway

Fix temperature and seeds where possible, average multiple samples, and use paired tests on the same items. Report confidence intervals, not a single percentage.