With the introduction of Large Language Models (LLMs), the sub-field of Artificial Intelligence, i.e., Natural Language Processing (NLP), is significantly advancing and improving. LLMs, with their remarkable text interpretation and generation abilities, are getting popular every day. These models are pre-trained using massive volumes of web data, the very best examples of that are the well-known GPT 3.5 AND GPT 4 models. Though the information on which the models are trained, i.e., the corpus, is large and varied, it is way from ideal. It’s unfiltered and noisy and includes false information in addition to factual errors. The query emerges as to how LLMs distinguish between truth and untruth when presented with a knowledge corpus that comprises each.

In a recent study, a team of researchers from Latest York University, ETH Zurich and Boston University proposed that LLMs can cluster truthful text, constructing on the premise that these models might represent different agents or sources contributing to the training data. By calling it a ‘truthful persona’, the researchers have shared that this persona stands for a group of agents that, as a result of shared text creation characteristics, usually tend to generate accurate and trustworthy information.

For example, reputable and well-established sites like Science and Wikipedia incessantly use formal writing styles and provides factual information regularly. LLMs are in a position to offer real responses outside of the actual situations through which each agent produced the training data by modelling this truthful persona. The team has shared two primary observations to support the persona hypothesis, that are as follows.

1. Pre-generation Truthfulness Assessment: Even before a model generates a solution, it is possible to find out if it is going to be truthful. This means that depending on the situation and the source agent’s persona, the LLM can evaluate a response’s truthfulness.

2. Enhancement of Truthfulness by Fantastic-Tuning: When LLMs are fine-tuned using a group of factual facts, they turn out to be more truthful about each unrelated and directly connected issues. This means that the true persona’s impact allows the model to generalise truthfulness principles to a wide range of subjects.

The team has evaluated the association between personas and model honesty by utilizing an artificial environment and mathematical processes. Different agents on this controlled scenario imagine various things about each mathematical operator, depending on how truthful or mistaken their beliefs are. These agents’ equations enable LLMs to reinforce their capability to answer previously unknown operators accurately and successfully discern between true and false assertions. This achievement is barely possible if actors within the training data share a truthful generative process that allows the development of a truthful identity.

In conclusion, this study shows that LLMs can acquire abstract concepts like truthfulness by making use of the hierarchical structures included of their training data. These models can generalise their ability to discern between true and false information and generate appropriate replies across a broad range of topics by modelling a real persona, even when the source agents for these topics share attributes suggestive of sincerity.

Try the Paper. All credit for this research goes to the researchers of this project. Also, don’t forget to affix our 32k+ ML SubReddit, 40k+ Facebook Community, Discord Channel, and Email Newsletter, where we share the newest AI research news, cool AI projects, and more.

When you like our work, you’ll love our newsletter..

We’re also on Telegram and WhatsApp.