Editing Factual Knowledge in Language Models

Abstract

The factual knowledge acquired during pretraining and stored in the parameters of Language Models (LMs) can be useful in downstream tasks (e.g., question answering or textual inference). However, some facts can be incorrectly induced or become obsolete over time. We present KNOWLEDGEEDITOR, a method which can be used to edit this knowledge and, thus, fix 'bugs' or unexpected predictions without the need for expensive retraining or fine-tuning. Besides being computationally efficient, KNOWLEDGEEDITOR does not require any modifications in LM pretraining (e.g., the use of meta-learning). In our approach, we train a hyper-network with constrained optimization to modify a fact without affecting the rest of the knowledge; the trained hyper-network is then used to predict the weight update at test time. We show KNOWL-EDGEEDITOR's efficacy with two popular architectures and knowledge-intensive tasks: i) a BERT model fine-tuned for fact-checking, and ii) a sequence-to-sequence BART model for question answering. With our method, changing a prediction on the specific wording of a query tends to result in a consistent change in predictions also for its paraphrases. We show that this can be further encouraged by exploiting (e.g., automatically-generated) paraphrases during training. Interestingly, our hyper-network can be regarded as a 'probe' revealing which components need to be changed to manipulate factual knowledge; our analysis shows that the updates tend to be concentrated on a small subset of components. 1

Fast Model Editing at Scale

“…A simple approach to making such edits is additional fine-tuning with a new label on the single example to be corrected. Yet fine-tuning on a single example tends to overfit, even when constraining the distance between the pre-and post-fine-tuning parameters (Zhu et al, 2020;Cao et al, 2021). This overfitting leads to failures of both locality and generality.…”

Locating and Editing Factual Associations in GPT

“…Our evaluations confirm a distinction between generalized knowing at the early MLP site and rote saying at the late selfattention site (Section 5.3). Furthermore, when compared to fine-tuning (Zhu et al, 2020) and meta-learning (Mitchell et al, 2021;De Cao et al, 2021), our benchmarks find that the explicitly localized ROME method avoids both generalization and specificity failures seen in other knowledge editing approaches, outperforming state-of-the-art opaque methods even at billion-parameter scale (Section 5.4).…”

“…Despite increasing adoption of this architecture, their knowledge representation remains under-explored. Research has been done for masked models (Petroni et al, 2019;Jiang et al, 2020;Elazar et al, 2021a;Geva et al, 2021;Dai et al, 2021;De Cao et al, 2021), but GPT's architectural differences (e.g., unidirectional attention, generation capabilities) provide an opportunity for new insights.…”

Time-Aware Language Models as Temporal Knowledge Bases

“…A related line of work has explored editing neural predictions after training given a dataset of revised input and output pairs (Sinitsin et al, 2020;Zhu et al, 2020;Cao et al, 2021). Here we introduce a different setting where we have access to new unlabeled text after model training, which must be used implicitly to update the factual predictions of the model.…”