What Matters in Memorizing and Recalling Facts? Multifaceted Benchmarks for Knowledge Probing in Language Models

TL;DR

This paper introduces BELIEFs, a multifaceted benchmark for evaluating the factual knowledge recall of language models, addressing limitations of existing methods and analyzing factors influencing memorization and recall abilities.

Contribution

It presents a new comprehensive benchmark, BELIEFs, and a large diverse dataset, MyriadLAMA, for evaluating knowledge recall in language models from multiple perspectives.

Findings

BELIEFs effectively evaluate PLMs' accuracy, consistency, and reliability.

Model size, training data, and tuning strategies significantly impact knowledge recall.

Prompt-based probing has limitations in fully assessing factual knowledge.

Abstract

Language models often struggle with handling factual knowledge, exhibiting factual hallucination issue. This makes it vital to evaluate the models' ability to recall its parametric knowledge about facts. In this study, we introduce a knowledge probing benchmark, BELIEF(ICL), to evaluate the knowledge recall ability of both encoder- and decoder-based pre-trained language models (PLMs) from diverse perspectives. BELIEFs utilize a multi-prompt dataset to evaluate PLM's accuracy, consistency, and reliability in factual knowledge recall. To enable a more reliable evaluation with BELIEFs, we semi-automatically create MyriadLAMA, which has massively diverse prompts. We validate the effectiveness of BELIEFs in comprehensively evaluating PLM's knowledge recall ability on diverse PLMs, including recent large language models (LLMs). We then investigate key factors in memorizing and recalling facts in PLMs, such as model size, pretraining strategy and corpora, instruction-tuning process and in-context learning settings. Finally, we reveal the limitation of the prompt-based knowledge probing. The MyriadLAMA is publicized.