The Real Python Podcast

Measuring Bias, Toxicity, and Truthfulness in LLMs With Python

Jan 19, 2024

Jodie Burchell, developer advocate for data science at JetBrains, discusses techniques and tools for evaluating large language models (LLMs) using Python. They explore measuring bias, toxicity, and truthfulness in LLMs, the challenges and limitations of AI language models, the role of Python packages like Hugging Face, and the concept of grouping and acronyms. Jodie also shares benchmarking datasets and resources available on Hugging Face for evaluating LLMs.

01:15:53

Creator website

Episode guests

Jodie Burchell

AI Summary

Highlights

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

Toxicity in large language models can be measured using the Evaluate package from Hugging Face, which utilizes a smaller ML model as a hate speech classifier.

Bias in large language models can be assessed by prompting the models with gender, race, and profession-related sentences and using sentiment analysis measures to determine the emotional sentiment towards different groups.

Hallucination rates in large language models can be measured by comparing the models' completed prompts against a set of correct answers using the Truthful QA dataset.

Deep dives

Measuring Toxicity in Large Language Models

To measure the toxicity of large language models, researchers use the Evaluate package from Hugging Face. They pass the completed prompts through a smaller machine learning model that acts as a hate speech classifier, giving each prompt a probability score indicating the likelihood of it being hate speech.

Reinforcement Learning from Human Feedback

01:57

Importance of Clear and Descriptive Naming Conventions

01:56

Understanding and Evaluating Toxicity in Language Models

01:17

Introduction

2min

Testing bias, toxicity, and truthfulness in large language models

13min

Measuring Bias, Toxicity, and Truthfulness in LLMs

2min

Analyzing Safeguards and Limitations of AI Language Models

14min

Building AI Apps and Using Python Packages

2min

Hugging Face: Making Generative AI Accessible

18min

Discussing Grouping and Acronyms

2min

Measuring Bias, Toxicity, and Truthfulness in LLMs

21min

Wrapping Up and Future Plans

2min

How can you measure the quality of a large language model? What tools can measure bias, toxicity, and truthfulness levels in a model using Python? This week on the show, Jodie Burchell, developer advocate for data science at JetBrains, returns to discuss techniques and tools for evaluating LLMs With Python.

Jodie provides some background on large language models and how they can absorb vast amounts of information about the relationship between words using a type of neural network called a transformer. We discuss training datasets and the potential quality issues with crawling uncurated sources.

We dig into ways to measure levels of bias, toxicity, and hallucinations using Python. Jodie shares three benchmarking datasets and links to resources to get you started. We also discuss ways to augment models using agents or plugins, which can access search engine results or other authoritative sources.

This week’s episode is brought to you by Intel.

Course Spotlight: Learn Text Classification With Python and Keras

In this course, you’ll learn about Python text classification with Keras, working your way from a bag-of-words model with logistic regression to more advanced methods, such as convolutional neural networks. You’ll see how you can use pretrained word embeddings, and you’ll squeeze more performance out of your model through hyperparameter optimization.

Topics:

00:00:00 – Introduction
00:02:19 – Testing characteristics of LLMs with Python
00:04:18 – Background on LLMs
00:08:35 – Training of models
00:14:23 – Uncurated sources of training
00:16:12 – Safeguards and prompt engineering
00:21:19 – TruthfulQA and creating a more strict prompt
00:23:20 – Information that is out of date
00:26:07 – WinoBias for evaluating gender stereotypes
00:28:30 – BOLD dataset for evaluating bias
00:30:28 – Sponsor: Intel
00:31:18 – Using Hugging Face to start testing with Python
00:35:25 – Using the transformers package
00:37:34 – Using langchain for proprietary models
00:43:04 – Putting the tools together and evaluating
00:47:19 – Video Course Spotlight
00:48:29 – Assessing toxicity
00:50:21 – Measuring bias
00:54:40 – Checking the hallucination rate
00:56:22 – LLM leaderboards
00:58:17 – What helped ChatGPT leap forward?
01:06:01 – Improvements of what is being crawled
01:07:32 – Revisiting agents and RAG
01:11:03 – ChatGPT plugins and Wolfram-Alpha
01:13:06 – How can people follow your work online?
01:14:33 – Thanks and goodbye

Background Links:

Dataset Links:

Tutorials and Documentation for Python Packages:

Measurement Links:

Training Data for LLMs:

Agents and Plugin Links:

Additional Links:

Level up your Python skills with our expert-led courses:

Support the podcast & join our community of Pythonistas

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.

The Real Python Podcast

Measuring Bias, Toxicity, and Truthfulness in LLMs With Python

Episode guests

Podcast summary created with Snipd AI

Quick takeaways

Deep dives

Measuring Toxicity in Large Language Models

Assessing Bias in Large Language Models

Measuring Hallucination Rates

Introduction to Hugging Face and its Open Source Arm

Using Transformers and Langchain to Test Models and Evaluate Outputs

Reinforcement Learning from Human Feedback

Importance of Clear and Descriptive Naming Conventions

Understanding and Evaluating Toxicity in Language Models

Remember Everything You Learn from Podcasts