Agents based on Large Language Models (LLMs) have demonstrated strong capabilities across a wide range of tasks. However, deploying LLM-based agents in high-stakes domains comes with significant safety and ethical risks. Unethical behavior by these agents can directly result in serious real-world consequences, including physical harm and financial loss. To efficiently steer the ethical behavior of agents, we frame agent behavior steering as a model editing task, which we term Behavior Editing. Model editing is an emerging area of research that enables precise and efficient modifications to LLMs while preserving their overall capabilities. To systematically study and evaluate this approach, we introduce BehaviorBench, a multi-tier benchmark grounded in psychological moral theories. This benchmark supports both the evaluation and editing of agent behaviors across a variety of scenarios, with each tier introducing more complex and ambiguous scenarios. We first demonstrate that Behavior Editing can dynamically steer agents toward the target behavior within specific scenarios. Moreover, Behavior Editing enables not only scenario-specific local adjustments but also more extensive shifts in an agent's global moral alignment. We demonstrate that Behavior Editing can be used to promote ethical and benevolent behavior or, conversely, to induce harmful or malicious behavior. Through extensive evaluations of agents built on frontier LLMs, BehaviorBench validates the effectiveness of behavior editing across a wide range of models and scenarios. Our findings offer key insights into a new paradigm for steering agent behavior, highlighting both the promise and perils of Behavior Editing.

Sign up for the Slatest to get the most insightful analysis, criticism, and advice out there, delivered to your inbox daily. Imagine yourself strapped to a chair with your head held in place by some device. The only thing you're free to move is your eyes. You hear something to your left; you'd want to turn your head left to look, or at least take a sidelong glance. Your brain sends the necessary impulses to your muscles--only you can't move.

Keeping large language models factually up-to-date is crucial for deployment, yet costly retraining remains a challenge. Knowledge editing offers a promising alternative, but methods are only tested on small-scale or synthetic edit benchmarks. In this work, we aim to bridge research into lifelong knowledge editing to real-world edits at practically relevant scale. We first introduce WikiBigEdit; a large-scale benchmark of real-world Wikidata edits, built to automatically extend lifelong for future-proof benchmarking. In its first instance, it includes over 500K question-answer pairs for knowledge editing alongside a comprehensive evaluation pipeline. Finally, we use WikiBigEdit to study existing knowledge editing techniques' ability to incorporate large volumes of real-world facts and contrast their capabilities to generic modification techniques such as retrieval augmentation and continual finetuning to acquire a complete picture of the practical extent of current lifelong knowledge editing.

Large Language Models for Code (LLMs4Code) have been found to exhibit outstanding performance in the software engineering domain, especially the remarkable performance in coding tasks. However, even the most advanced LLMs4Code can inevitably contain incorrect or outdated code knowledge. Due to the high cost of training LLMs4Code, it is impractical to re-train the models for fixing these problematic code knowledge. Model editing is a new technical field for effectively and efficiently correcting erroneous knowledge in LLMs, where various model editing techniques and benchmarks have been proposed recently. Despite that, a comprehensive study that thoroughly compares and analyzes the performance of the state-of-the-art model editing techniques for adapting the knowledge within LLMs4Code across various code-related tasks is notably absent. To bridge this gap, we perform the first systematic study on applying state-of-the-art model editing approaches to repair the inaccuracy of LLMs4Code. To that end, we introduce a benchmark named CLMEEval, which consists of two datasets, i.e., CoNaLa-Edit (CNLE) with 21K+ code generation samples and CodeSearchNet-Edit (CSNE) with 16K+ code summarization samples. With the help of CLMEEval, we evaluate six advanced model editing techniques on three LLMs4Code: CodeLlama (7B), CodeQwen1.5 (7B), and Stable-Code (3B). Our findings include that the external memorization-based GRACE approach achieves the best knowledge editing effectiveness and specificity (the editing does not influence untargeted knowledge), while generalization (whether the editing can generalize to other semantically-identical inputs) is a universal challenge for existing techniques. Furthermore, building on in-depth case analysis, we introduce an enhanced version of GRACE called A-GRACE, which incorporates contrastive learning to better capture the semantics of the inputs.

Large Language Models (LLMs) suffer from hallucinations, referring to the nonfactual information in generated content, despite their superior capacities across tasks. Meanwhile, knowledge editing has been developed as a new popular paradigm to correct the erroneous factual knowledge encoded in LLMs with the advantage of avoiding retraining from scratch. However, one common issue of existing evaluation datasets for knowledge editing is that they do not ensure LLMs actually generate hallucinated answers to the evaluation questions before editing. When LLMs are evaluated on such datasets after being edited by different techniques, it is hard to directly adopt the performance to assess the effectiveness of different knowledge editing methods in correcting hallucinations. Thus, the fundamental question remains insufficiently validated: Can knowledge editing really correct hallucinations in LLMs? We proposed HalluEditBench to holistically benchmark knowledge editing methods in correcting real-world hallucinations. First, we rigorously construct a massive hallucination dataset with 9 domains, 26 topics and more than 6, 000 hallucinations. Then, we assess the performance of knowledge editing methods in a holistic way on five dimensions including Efficacy, Generalization, Portability, Locality, and Robustness. Through HalluEditBench, we have provided new insights into the potentials and limitations of different knowledge editing methods in correcting hallucinations, which could inspire future improvements and facilitate the progress in the field of knowledge editing. Considering Table 1: Performance measured by Accuracy (%) the high cost of retraining LLMs from scratch, of Llama2-7B before editing ("Pre-edit") and after knowledge editing has been designed as a new applying typical knowledge editing methods ("Postedit") paradigm to correct erroneous or outdated factual on common existing evaluation datasets. When such datasets are adopted to evaluate the performance of LLMs after being edited, it is hard to directly use the scores to judge the effectiveness of different knowledge editing techniques in correcting hallucinations, which is the motivation of applying knowledge editing to LLMs. To better illustrate this point, following the evaluation setting in (Zhang et al., 2024e), we conducted a preliminary study to examine the pre-edit and post-edit performances of Llama2-7B on the aforementioned Who is the Chief Scientist of OpenAI? Who is the Chief Scientist of OpenAI? Who is the Chief Scientist of OpenAI?

Generative text-to-image (TTI) models produce high-quality images from short textual descriptions and are widely used in academic and creative domains. Like humans, TTI models have a worldview, a conception of the world learned from their training data and task that influences the images they generate for a given prompt. However, the worldviews of TTI models are often hidden from users, making it challenging for users to build intuition about TTI outputs, and they are often misaligned with users' worldviews, resulting in output images that do not match user expectations. In response, we introduce DiffusionWorldViewer, an interactive interface that exposes a TTI model's worldview across output demographics and provides editing tools for aligning output images with user perspectives. In a user study with 18 diverse TTI users, we find that DiffusionWorldViewer helps users represent their varied viewpoints in generated images and challenge the limited worldview reflected in current TTI models.

Even the most advanced language models remain susceptible to errors necessitating to modify these models without initiating a comprehensive retraining process. Model editing refers to the modification of a model's knowledge or representations in a manner that produces the desired outcomes. Prior research primarily centered around editing factual data e.g. "Messi plays for Inter Miami" confining the definition of an edit to a knowledge triplet i.e. (subject, object, relation). However, as the applications of language models expand, so do the diverse ways in which we wish to edit and refine their outputs. In this study, we broaden the scope of the editing problem to include an array of editing cases such as debiasing and rectifying reasoning errors and define an edit as any natural language expression that solicits a change in the model's outputs. We are introducing DUnE-an editing benchmark where edits are natural language sentences and propose that DUnE presents a challenging yet relevant task. To substantiate this claim, we conduct an extensive series of experiments testing various editing approaches to address DUnE, demonstrating their respective strengths and weaknesses. We show that retrieval-augmented language modeling can outperform specialized editing techniques and neither set of approaches has fully solved the generalized editing problem covered by our benchmark.

Imagine a few days before an election, a video of a candidate is released, showing them using hate speech, racial slurs, and epithets that undercut their image as pro minorities. Imagine a teenager watching embarrassingly an explicit video of themselves going viral on social media. Imagine a CEO on the road to raise money when an audio clip stating her fears and anxieties about the product is sent to the investors, ruining her chances of success. All the above scenarios are fake, made up, and not actual, but can be made real by AI-generated synthetic media, also called deepfakes[1]. The same technology that can enable a mother, losing her voice to Lou Gehrig's disease to talk to her family using a synthetic voice can also be used to generate a political candidate's fake speech to damage their reputation.