Estimating Knowledge in Large Language Models Without Generating a Single Token

TL;DR

This paper introduces KEEN, a lightweight probe that estimates a large language model's knowledge about an entity solely from its internal representations, without generating any text, enabling pre-response evaluation.

Contribution

The work presents KEEN, a novel internal probe that predicts model knowledge and factuality without token generation, improving evaluation efficiency and interpretability.

Findings

KEEN correlates with QA accuracy and factuality metrics.

KEEN reflects knowledge changes after fine-tuning.

A variant of KEEN offers interpretability with token clusters.

Abstract

To evaluate knowledge in large language models (LLMs), current methods query the model and then evaluate its generated responses. In this work, we ask whether evaluation can be done before the model has generated any text. Concretely, is it possible to estimate how knowledgeable a model is about a certain entity, only from its internal computation? We study this question with two tasks: given a subject entity, the goal is to predict (a) the ability of the model to answer common questions about the entity, and (b) the factuality of open-ended responses generated by the model about the entity. Experiments with a variety of LLMs show that KEEN, a simple probe trained over internal subject representations, succeeds at both tasks - correlating with both the QA accuracy of the model per-subject and FActScore, a recent factuality metric in open-ended generation. Moreover, KEEN naturally aligns with the model's hedging behavior and faithfully reflects changes in the model's knowledge after fine-tuning. Lastly, we show a more interpretable yet equally performant variant of KEEN, which highlights a small set of tokens indicative of clusters and gaps in the model's knowledge. Being simple and lightweight, KEEN can be leveraged to guide decisions such as when it is appropriate to apply further training or augment queries with retrieval.