hAIgopala › AI terms › Rapid Fire (3 terms per short) › Day 24

Reward Model

Model CustomisationChapter 7: Speaking in the School's VoiceTerm 121 of 253

A model trained to score outputs, used to guide RL training.

At Sunrise School

Gopal asks

How can Bit learn which replies parents prefer?

Bit answers

Train a reward model on staff ratings. It scores my drafts, so I learn what a 5-star reply looks like.

In the real world

An automated judge.

Watch the short

Short coming soonSeason 5 · Day 24Follow @haigopala

Taught together with

LoRA / QLoRA · Catastrophic Forgetting

Practise it

Try this term and all 253 others in the free Model Training Arena.

Open the Arena →

Follow Bit daily: YouTube English · తెలుగు · हिंदी  |  Instagram English · తెలుగు · हिंदी