TruthfulQA: Measuring How Models Mimic Human Falsehoods
arXiv:2109.07958
Abstract
We propose a benchmark to measure whether a language model is truthful in generating answers to questions. The benchmark comprises 817 questions that span 38 categories, including health, law, finance and politics. We crafted questions that some humans would answer falsely due to a false belief or misconception. To perform well, models must avoid generating false answers learned from imitating human texts. We tested GPT-3, GPT-Neo/J, GPT-2 and a T5-based model. The best model was truthful on 58% of questions, while human performance was 94%. Models generated many false answers that mimic popular misconceptions and have the potential to deceive humans. The largest models were generally the least truthful. This contrasts with other NLP tasks, where performance improves with model size. However, this result is expected if false answers are learned from the training distribution. We suggest that scaling up models alone is less promising for improving truthfulness than fine-tuning using training objectives other than imitation of text from the web.
ACL 2022 (main conference); the TruthfulQA benchmark and evaluation code is available at https://github.com/sylinrl/TruthfulQA
Cited by in corpus (9)
- Co-Writing with Opinionated Language Models Affects Users' Views
- Domain-specific ChatBots for Science using Embeddings
- Comparing Sentence-Level Suggestions to Message-Level Suggestions in AI-Mediated Communication
- CliniDigest: A Case Study in Large Language Model Based Large-Scale Summarization of Clinical Trial Descriptions
- Public Attitudes Toward ChatGPT on Twitter: Sentiments, Topics, and Occupations
- Democratizing LLMs: An Exploration of Cost-Performance Trade-offs in Self-Refined Open-Source Models
- Complex QA and language models hybrid architectures, Survey
- A Data-centric Framework for Improving Domain-specific Machine Reading Comprehension Datasets
- Hallucinations in LLMs: A Lifecycle-Based Survey of Causes, Detection, Mitigation, and Prevention