# Lab 3, 2.2 Reward Model

**URL:** <https://community.deeplearning.ai/t/lab-3-2-2-reward-model/536980>\
**Category:** Generative AI with Large Language Models\
**Tags:** week-module-3\
**Created:** [January 6, 2024, 9:33am UTC](https://community.deeplearning.ai/t/lab-3-2-2-reward-model/536980 "2024-01-06T09:33:37Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![OLOWOBOKO\_EMMANUEL](https://avatars.discourse-cdn.com/v4/letter/o/94ad74/32.png) [@OLOWOBOKO\_EMMANUEL](https://community.deeplearning.ai/u/OLOWOBOKO_EMMANUEL)\
**Post date:** [January 6, 2024, 9:33am UTC](https://community.deeplearning.ai/t/lab-3-2-2-reward-model/536980/1 "2024-01-06T09:33:37Z")

</div>

How can I extend the reward model beyond checking for toxicity in the generated output to other specific criteria or options? In lab 3 2.2, the approach used is sentiment analysis using “Meta AI’s RoBERTa-based hate speech model” which gives a higher reward if there is higher a chance of getting class nothate as an output.