Knowledge editing techniques, aiming to efficiently modify a minor proportion of knowledge in large language models (LLMs) without negatively impacting performance across other inputs, have garnered widespread attention. However, existing methods predominantly rely on memorizing the updated knowledge, impeding LLMs from effectively combining the new knowledge with their inherent knowledge when answering questions. To this end, we propose a Learning to Edit (LTE) framework, focusing on teaching LLMs to apply updated knowledge into input questions, inspired by the philosophy of "Teach a man to fish." LTE features a two-phase process: (i) the Alignment Phase, which fine-tunes LLMs on a meticulously curated parallel dataset to make reliable, in-scope edits while preserving out-of-scope information and linguistic proficiency; and (ii) the Inference Phase, which employs a retrieval-based mechanism for real-time and mass knowledge editing. By comparing our approach with seven advanced baselines across four popular knowledge editing benchmarks and two LLM architectures, we demonstrate LTE's superiority in knowledge editing performance, robustness in both batch and sequential editing, minimal interference on general tasks, and rapid editing speeds. The data and code are available at https://github.com/YJiangcm/LTE.
Recent work has showcased the powerful capability of large language models (LLMs) in recalling knowledge and reasoning. However, the reliability of LLMs in combining these two capabilities into reasoning through multi-hop facts has not been widely explored. This paper systematically investigates the possibilities for LLMs to utilize shortcuts based on direct connections between the initial and terminal entities of multi-hop knowledge. We first explore the existence of factual shortcuts through Knowledge Neurons, revealing that: (i) the strength of factual shortcuts is highly correlated with the frequency of co-occurrence of initial and terminal entities in the pre-training corpora; (ii) few-shot prompting leverage more shortcuts in answering multi-hop questions compared to chain-of-thought prompting. Then, we analyze the risks posed by factual shortcuts from the perspective of multi-hop knowledge editing. Analysis shows that approximately 20% of the failures are attributed to shortcuts, and the initial and terminal entities in these failure instances usually have higher co-occurrences in the pre-training corpus. Finally, we propose erasing shortcut neurons to mitigate the associated risks and find that this approach significantly reduces failures in multiple-hop knowledge editing caused by shortcuts.
Knowledge editing aims to rectify inaccuracies in large language models (LLMs) without costly retraining for outdated or erroneous knowledge. However, current knowledge editing methods primarily focus on single editing, failing to meet the requirements for lifelong editing. This study reveals a performance degradation encountered by knowledge editing in lifelong editing, characterized by toxicity buildup and toxicity flash, with the primary cause identified as pattern unmatch. We introduce a knowledge editing approach named Wise-Layer Knowledge Editor (WilKE), which selects editing layer based on the pattern matching degree of editing knowledge across different layers in language models. Experimental results demonstrate that, in lifelong editing, WilKE exhibits an average improvement of 46.2% and 67.8% on editing GPT2-XL and GPT-J relative to state-of-the-art knowledge editing methods.
Knowledge editing (KE) aims to efficiently and precisely modify the behavior of large language models (LLMs) to update specific knowledge without negatively influencing other knowledge. Current research primarily focuses on white-box LLMs editing, overlooking an important scenario: black-box LLMs editing, where LLMs are accessed through interfaces and only textual output is available. In this paper, we first officially introduce KE on black-box LLMs and then propose a comprehensive evaluation framework to overcome the limitations of existing evaluations that are not applicable to black-box LLMs editing and lack comprehensiveness. To tackle privacy leaks of editing data and style over-editing in current methods, we introduce a novel postEdit framework, resolving privacy concerns through downstream post-processing and maintaining textual style consistency via fine-grained editing to original responses. Experiments and analysis on two benchmarks demonstrate that postEdit outperforms all baselines and achieves strong generalization, especially with huge improvements on style retention (average $+20.82\%\uparrow$).
Despite their exceptional capabilities, large language models (LLMs) are prone to generating unintended text due to false or outdated knowledge. Given the resource-intensive nature of retraining LLMs, there has been a notable increase in the development of knowledge editing. However, current approaches and evaluations rarely explore the perturbation of editing on neighboring knowledge. This paper studies whether updating new knowledge to LLMs perturbs the neighboring knowledge encapsulated within them. Specifically, we seek to figure out whether appending a new answer into an answer list to a factual question leads to catastrophic forgetting of original correct answers in this list, as well as unintentional inclusion of incorrect answers. A metric of additivity is introduced and a benchmark dubbed as Perturbation Evaluation of Appending Knowledge (PEAK) is constructed to evaluate the degree of perturbation to neighboring knowledge when appending new knowledge. Besides, a plug-and-play framework termed Appending via Preservation and Prevention (APP) is proposed to mitigate the neighboring perturbation by maintaining the integrity of the answer list. Experiments demonstrate the effectiveness of APP coupling with four editing methods on four LLMs. The code and data are available at https://github.com/mjy1111/PEAK.
How to edit the knowledge in multi-step reasoning has become the major challenge in the knowledge editing (KE) of large language models (LLMs). The difficulty arises because the hallucinations of LLMs during multi-step reasoning often lead to incorrect use of new knowledge and incorrect answers. To address this issue, we design decoding constraints to "regulate" LLMs' reasoning, enhancing logical coherence when incorporating new knowledge. We propose a new KE framework: DEEPEDIT (Depth-first Search-based Constrained Decoding for Knowledge Editing), which enhances LLMs's ability to generate coherent reasoning chains with new knowledge through depth-first search. Our search selects the most important knowledge that satisfies our constraints as the reasoning step to efficiently increase the reasoning depth. In addition to DEEPEDIT, we propose two new KE benchmarks: MQUAKE-2002 and MQUAKE-HARD, which provide more precise and challenging assessments of KE approaches. Qualitatively, DEEPEDIT enables LLMs to produce succinct and coherent reasoning chains involving new knowledge. Quantitatively, it yields significant improvements on multiple KE benchmarks.
Language models learn a great quantity of factual information during pretraining, and recent work localizes this information to specific model weights like mid-layer MLP weights. In this paper, we find that we can change how a fact is stored in a model by editing weights that are in a different location than where existing methods suggest that the fact is stored. This is surprising because we would expect that localizing facts to specific model parameters would tell us where to manipulate knowledge in models, and this assumption has motivated past work on model editing methods. Specifically, we show that localization conclusions from representation denoising (also known as Causal Tracing) do not provide any insight into which model MLP layer would be best to edit in order to override an existing stored fact with a new one. This finding raises questions about how past work relies on Causal Tracing to select which model layers to edit. Next, we consider several variants of the editing problem, including erasing and amplifying facts. For one of our editing problems, editing performance does relate to localization results from representation denoising, but we find that which layer we edit is a far better predictor of performance. Our results suggest, counterintuitively, that better mechanistic understanding of how pretrained language models work may not always translate to insights about how to best change their behavior. Our code is available at https://github.com/google/belief-localization
Large Language Models (LLMs) have shown extraordinary capabilities in understanding and generating text that closely mirrors human communication. However, a primary limitation lies in the significant computational demands during training, arising from their extensive parameterization. This challenge is further intensified by the dynamic nature of the world, necessitating frequent updates to LLMs to correct outdated information or integrate new knowledge, thereby ensuring their continued relevance. Note that many applications demand continual model adjustments post-training to address deficiencies or undesirable behaviors. There is an increasing interest in efficient, lightweight methods for on-the-fly model modifications. To this end, recent years have seen a burgeoning in the techniques of knowledge editing for LLMs, which aim to efficiently modify LLMs' behaviors within specific domains while preserving overall performance across various inputs. In this paper, we first define the knowledge editing problem and then provide a comprehensive review of cutting-edge approaches. Drawing inspiration from educational and cognitive research theories, we propose a unified categorization criterion that classifies knowledge editing methods into three groups: resorting to external knowledge, merging knowledge into the model, and editing intrinsic knowledge. Furthermore, we introduce a new benchmark, KnowEdit, for a comprehensive empirical evaluation of representative knowledge editing approaches. Additionally, we provide an in-depth analysis of knowledge location, which can give a deeper understanding of the knowledge structures inherent within LLMs. Finally, we discuss several potential applications of knowledge editing, outlining its broad and impactful implications.
Multi-hop question answering (MQA) is one of the challenging tasks to evaluate machine's comprehension and reasoning abilities, where large language models (LLMs) have widely achieved the human-comparable performance. Due to the dynamics of knowledge facts in real world, knowledge editing has been explored to update model with the up-to-date facts while avoiding expensive re-training or fine-tuning. Starting from the edited fact, the updated model needs to provide cascading changes in the chain of MQA. The previous art simply adopts a mix-up prompt to instruct LLMs conducting multiple reasoning tasks sequentially, including question decomposition, answer generation, and conflict checking via comparing with edited facts. However, the coupling of these functionally-diverse reasoning tasks inhibits LLMs' advantages in comprehending and answering questions while disturbing them with the unskilled task of conflict checking. We thus propose a framework, Programmable knowledge editing for Multi-hop Question Answering (PokeMQA), to decouple the jobs. Specifically, we prompt LLMs to decompose knowledge-augmented multi-hop question, while interacting with a detached trainable scope detector to modulate LLMs behavior depending on external conflict signal. The experiments on three LLM backbones and two benchmark datasets validate our superiority in knowledge editing of MQA, outperforming all competitors by a large margin in almost all settings and consistently producing reliable reasoning process.
Knowledge represented in Large Language Models (LLMs) is quite often incorrect and can also become obsolete over time. Updating knowledge via fine-tuning is computationally resource-hungry and not reliable, and so knowledge editing (KE) has developed as an effective and economical alternative to inject new knowledge or to fix factual errors in LLMs. Although there has been considerable interest in this area, current KE research exclusively focuses on the monolingual setting, typically in English. However, what happens if the new knowledge is supplied in one language, but we would like to query the LLM in a different language? To address the problem of multilingual knowledge editing, we propose Retrieval-augmented Multilingual Knowledge Editor (ReMaKE) to update new knowledge in LLMs. ReMaKE can perform model-agnostic knowledge editing in multilingual settings. ReMaKE concatenates the new knowledge retrieved from a multilingual knowledge base with prompts. Our experimental results show that ReMaKE outperforms baseline knowledge editing methods by a significant margin and is the first KE method to work in a multilingual setting. We provide our multilingual knowledge editing dataset (MzsRE) in 12 languages, which along with code, and additional project information is available at https://github.com/Vicky-Wil/ReMaKE.
Identifying important neurons for final predictions is essential for understanding the mechanisms of large language models. Due to computational constraints, current attribution techniques struggle to operate at neuron level. In this paper, we propose a static method for pinpointing significant neurons. Compared to seven other methods, our approach demonstrates superior performance across three metrics. Additionally, since most static methods typically only identify "value neurons" directly contributing to the final prediction, we propose a method for identifying "query neurons" which activate these "value neurons". Finally, we apply our methods to analyze six types of knowledge across both attention and feed-forward network (FFN) layers. Our method and analysis are helpful for understanding the mechanisms of knowledge storage and set the stage for future research in knowledge editing. The code is available on https://github.com/zepingyu0512/neuron-attribution.
The imperative task of revising or updating the knowledge stored within large language models arises from two distinct sources: intrinsic errors inherent in the model which should be corrected and outdated knowledge due to external shifts in the real world which should be updated. Prevailing efforts in model editing conflate these two distinct categories of edits arising from distinct reasons and directly modify the original knowledge in models into new knowledge. However, we argue that preserving the model's original knowledge remains pertinent. Specifically, if a model's knowledge becomes outdated due to evolving worldly dynamics, it should retain recollection of the historical knowledge while integrating the newfound knowledge. In this work, we introduce the task of Temporal Knowledge Editing (TKE) and establish a benchmark AToKe (Assessment of TempOral Knowledge Editing) to evaluate current model editing methods. We find that while existing model editing methods are effective at making models remember new knowledge, the edited model catastrophically forgets historical knowledge. To address this gap, we propose a simple and general framework termed Multi-Editing with Time Objective (METO) for enhancing existing editing models, which edits both historical and new knowledge concurrently and optimizes the model's prediction for the time of each fact. Our assessments demonstrate that while AToKe is still difficult, METO maintains the effectiveness of learning new knowledge and meanwhile substantially improves the performance of edited models on utilizing historical knowledge.
Modern language models have the capacity to store and use immense amounts of knowledge about real-world entities, but it remains unclear how to update such knowledge stored in model parameters. While prior methods for updating knowledge in LMs successfully inject atomic facts, updated LMs fail to make inferences based on injected facts. In this work, we demonstrate that a context distillation-based approach can both impart knowledge about entities and propagate that knowledge to enable broader inferences. Our approach consists of two stages: transfer set generation and distillation on the transfer set. We first generate a transfer set by prompting a language model to generate continuations from the entity definition. Then, we update the model parameters so that the distribution of the LM (the student) matches the distribution of the LM conditioned on the definition (the teacher) on the transfer set. Our experiments demonstrate that this approach is more effective at propagating knowledge updates than fine-tuning and other gradient-based knowledge-editing methods. Moreover, it does not compromise performance in other contexts, even when injecting the definitions of up to 150 entities at once.
The potential of using a large language model (LLM) as a knowledge base (KB) has sparked significant interest. To manage the knowledge acquired by LLMs, we need to ensure that the editing of learned facts respects internal logical constraints, which are known as dependency of knowledge. Existing work on editing LLMs has partially addressed the issue of dependency, when the editing of a fact should apply to its lexical variations without disrupting irrelevant ones. However, they neglect the dependency between a fact and its logical implications. We propose an evaluation protocol with an accompanying question-answering dataset, DepEdit, that provides a comprehensive assessment of the editing process considering the above notions of dependency. Our protocol involves setting up a controlled environment in which we edit facts and monitor their impact on LLMs, along with their implications based on If-Then rules. Extensive experiments on DepEdit show that existing knowledge editing methods are sensitive to the surface form of knowledge, and that they have limited performance in inferring the implications of edited facts.
Knowledge Editing (KE) for modifying factual knowledge in Large Language Models (LLMs) has been receiving increasing attention. However, existing knowledge editing methods are entity-centric, and it is unclear whether this approach is suitable for a relation-centric perspective. To address this gap, this paper constructs a new benchmark named RaKE, which focuses on Relation based Knowledge Editing. In this paper, we establish a suite of innovative metrics for evaluation and conduct comprehensive experiments involving various knowledge editing baselines. We notice that existing knowledge editing methods exhibit the potential difficulty in their ability to edit relations. Therefore, we further explore the role of relations in factual triplets within the transformer. Our research results confirm that knowledge related to relations is not only stored in the FFN network but also in the attention layers. This provides experimental support for future relation-based knowledge editing methods.
Understanding the internal mechanisms by which multi-modal large language models (LLMs) interpret different modalities and integrate cross-modal representations is becoming increasingly critical for continuous improvements in both academia and industry. In this paper, we propose a novel method to identify key neurons for interpretability -- how multi-modal LLMs bridge visual and textual concepts for captioning. Our method improves conventional works upon efficiency and applied range by removing needs of costly gradient computation. Based on those identified neurons, we further design a multi-modal knowledge editing method, beneficial to mitigate sensitive words or hallucination. For rationale of our design, we provide theoretical assumption. For empirical evaluation, we have conducted extensive quantitative and qualitative experiments. The results not only validate the effectiveness of our methods, but also offer insightful findings that highlight three key properties of multi-modal neurons: sensitivity, specificity and causal-effect, to shed light for future research.
Recent studies have demonstrated that large language models (LLMs) store massive factual knowledge within their parameters. But existing LLMs are prone to hallucinate unintended text due to false or outdated knowledge. Since retraining LLMs is resource intensive, there has been a growing interest in the concept of model editing. Despite the emergence of benchmarks and approaches, these unidirectional editing and evaluation have failed to explore the reversal curse. Intuitively, if "The capital of France is" is edited to be a counterfact "London" within a model, then it should be able to naturally reason and recall the reverse fact, i.e., "London is the capital of" followed by "France" instead of "England". In this paper, we study bidirectional language model editing, aiming to provide rigorous model editing evaluation to assess if edited LLMs can recall the editing knowledge bidirectionally. A new evaluation metric of reversibility is introduced, and a benchmark dubbed as Bidirectional Assessment for Knowledge Editing (BAKE) is constructed to evaluate the reversibility of edited models in recalling knowledge in the reverse direction of editing. We surprisingly observe that while current editing methods and LLMs can effectively recall editing facts in the direction of editing, they suffer serious deficiencies when evaluated in the reverse direction. To mitigate the reversal curse, a method named Bidirectionally Inversible Relationship moDeling (BIRD) is proposed. A set of editing objectives that incorporate bidirectional relationships between subject and object into the updated model weights are designed. Experiments show that BIRD improves the performance of four representative LLMs of different sizes via question answering and judgement.
As the cost associated with fine-tuning Large Language Models (LLMs) continues to rise, recent research efforts have pivoted towards developing methodologies to edit implicit knowledge embedded within LLMs. Yet, there's still a dark cloud lingering overhead -- will knowledge editing trigger butterfly effect? since it is still unclear whether knowledge editing might introduce side effects that pose potential risks or not. This paper pioneers the investigation into the potential pitfalls associated with knowledge editing for LLMs. To achieve this, we introduce new benchmark datasets and propose innovative evaluation metrics. Our results underline two pivotal concerns: (1) Knowledge Conflict: Editing groups of facts that logically clash can magnify the inherent inconsistencies in LLMs-a facet neglected by previous methods. (2) Knowledge Distortion: Altering parameters with the aim of editing factual knowledge can irrevocably warp the innate knowledge structure of LLMs. Experimental results vividly demonstrate that knowledge editing might inadvertently cast a shadow of unintended consequences on LLMs, which warrant attention and efforts for future works. Code and data are available at https://github.com/zjunlp/PitfallsKnowledgeEditing.
Knowledge editing aims to change language models' performance on several special cases (i.e., editing scope) by infusing the corresponding expected knowledge into them. With the recent advancements in large language models (LLMs), knowledge editing has been shown as a promising technique to adapt LLMs to new knowledge without retraining from scratch. However, most of the previous studies neglect the multi-lingual nature of some main-stream LLMs (e.g., LLaMA, ChatGPT and GPT-4), and typically focus on monolingual scenarios, where LLMs are edited and evaluated in the same language. As a result, it is still unknown the effect of source language editing on a different target language. In this paper, we aim to figure out this cross-lingual effect in knowledge editing. Specifically, we first collect a large-scale cross-lingual synthetic dataset by translating ZsRE from English to Chinese. Then, we conduct English editing on various knowledge editing methods covering different paradigms, and evaluate their performance in Chinese, and vice versa. To give deeper analyses of the cross-lingual effect, the evaluation includes four aspects, i.e., reliability, generality, locality and portability. Furthermore, we analyze the inconsistent behaviors of the edited models and discuss their specific challenges. Data and codes are available at https://github.com/krystalan/Bi_ZsRE
Pre-trained language models (PLMs) contain vast amounts of factual knowledge, but how the knowledge is stored in the parameters remains unclear. This paper delves into the complex task of understanding how factual knowledge is stored in multilingual PLMs, and introduces the Architecture-adapted Multilingual Integrated Gradients method, which successfully localizes knowledge neurons more precisely compared to current methods, and is more universal across various architectures and languages. Moreover, we conduct an in-depth exploration of knowledge neurons, leading to the following two important discoveries: (1) The discovery of Language-Independent Knowledge Neurons, which store factual knowledge in a form that transcends language. We design cross-lingual knowledge editing experiments, demonstrating that the PLMs can accomplish this task based on language-independent neurons; (2) The discovery of Degenerate Knowledge Neurons, a novel type of neuron showing that different knowledge neurons can store the same fact. Its property of functional overlap endows the PLMs with a robust mastery of factual knowledge. We design fact-checking experiments, proving that the degenerate knowledge neurons can help the PLMs to detect wrong facts. Experiments corroborate these findings, shedding light on the mechanisms of factual knowledge storage in multilingual PLMs, and contribute valuable insights to the field. The code is available at https://github.com/heng840/AMIG.