Keywords: LLMs, Factuality, Uncertainty, Factual Knowledge, Pretraining
TL;DR: We propose a novel IDK objective which models uncertainty in a model's prediction as probability mass put on a special [IDK] token. Experimental results show that our method improves factual precision without significantly harming factual recall.
Abstract: Large Language Models are known to capture real-world knowledge, allowing them to excel in many downstream tasks. Despite recent advances, these models are still prone to what are commonly known as hallucinations, causing them to emit unwanted and factually incorrect text. In this work, we propose a novel calibration method that can be used to combat hallucinations.
We add a special [IDK] (“I Don't Know”) token to the model's vocabulary and introduce an objective function that shifts probability mass to the [IDK] token for incorrect predictions.
This approach allows the model to express uncertainty in its output explicitly.
We evaluate our proposed method across multiple model architectures and factual downstream tasks.
We find that models trained with our method are able to express uncertainty in places where they would previously make mistakes while suffering only a small loss of encoded knowledge. We further perform extensive ablation studies of multiple variations of our approach and provide a detailed analysis of the precision-recall tradeoff of our method.
Primary Area: Natural language processing
Submission Number: 18862
Loading