Large Language Models (LLMs) have been found to memorize and recite some of the textual sequences from their training set verbatim, raising broad concerns about privacy and copyright issues when using LLMs. This Textual Sequence Memorization (TSM) phenomenon leads to a high demand to regulate LLM output to prevent it from generating certain memorized text to meet user requirements. However, our empirical study reveals that existing methods for TSM erasure fail to forget massive memorized samples without substantially jeopardizing the model utility. To achieve a better trade-off between the effectiveness of TSM erasure and model utility in LLMs, our paper proposes a new framework based on Entropy Maximization with Selective Optimization (EMSO), where the updated weights are chosen with a novel contrastive gradient metric without any participation of additional model or data. Our analysis shows that training with the entropy maximization loss has a more stable optimization process and better keeps model utility than existing methods. The contrastive gradient metric localizes the most influential weight for TSM erasure by taking both the gradient magnitude and direction into consideration. Extensive experiments across three model scales demonstrate that our method excels in handling large-scale forgetting requests while preserving model ability in language generation and reasoning.

In the pursuit of enhancing the efficacy and flexibility of interpretable, data-driven classification models, this work introduces a novel incorporation of user-defined preferences with Abstract Argumentation and Case-Based Reasoning (CBR). Specifically, we introduce Preference-Based Abstract Argumentation for Case-Based Reasoning (which we call AA-CBR-P), allowing users to define multiple approaches to compare cases with an ordering that specifies their preference over these comparison approaches. We prove that the model inherently follows these preferences when making predictions and show that previous abstract argumentation for case-based reasoning approaches are insufficient at expressing preferences over constituents of an argument. We then demonstrate how this can be applied to a real-world medical dataset sourced from a clinical trial evaluating differing assessment methods of patients with a primary brain tumour. We show empirically that our approach outperforms other interpretable machine learning models on this dataset.

In the rapidly evolving landscape of artificial intelligence, generative models such as Generative Adversarial Networks (GANs) and Diffusion Models have become cornerstone technologies, driving innovation in diverse fields from art creation to healthcare. Despite their potential, these models face the significant challenge of data memorization, which poses risks to privacy and the integrity of generated content. Among various metrics of memorization detection, our study delves into the memorization scores calculated from encoder layer embeddings, which involves measuring distances between samples in the embedding spaces. Particularly, we find that the memorization scores calculated from layer embeddings of Vision Transformers (ViTs) show an notable trend - the latter (deeper) the layer, the less the memorization measured. It has been found that the memorization scores from the early layers' embeddings are more sensitive to low-level memorization (e.g. colors and simple patterns for an image), while those from the latter layers are more sensitive to high-level memorization (e.g. semantic meaning of an image). We also observe that, for a specific model architecture, its degree of memorization on different levels of information is unique. It can be viewed as an inherent property of the architecture. Building upon this insight, we introduce a unique fingerprinting methodology. This method capitalizes on the unique distributions of the memorization score across different layers of ViTs, providing a novel approach to identifying models involved in generating deepfakes and malicious content. Our approach demonstrates a marked 30% enhancement in identification accuracy over existing baseline methods, offering a more effective tool for combating digital misinformation.

Large Language Models (LLMs) frequently memorize long sequences verbatim, often with serious legal and privacy implications. Much prior work has studied such verbatim memorization using observational data. To complement such work, we develop a framework to study verbatim memorization in a controlled setting by continuing pre-training from Pythia checkpoints with injected sequences. We find that (1) non-trivial amounts of repetition are necessary for verbatim memorization to happen; (2) later (and presumably better) checkpoints are more likely to verbatim memorize sequences, even for out-of-distribution sequences; (3) the generation of memorized sequences is triggered by distributed model states that encode high-level features and makes important use of general language modeling capabilities. Guided by these insights, we develop stress tests to evaluate unlearning methods and find they often fail to remove the verbatim memorized information, while also degrading the LM. Overall, these findings challenge the hypothesis that verbatim memorization stems from specific model weights or mechanisms. Rather, verbatim memorization is intertwined with the LM's general capabilities and thus will be very difficult to isolate and suppress without degrading model quality.

This short paper presents preliminary research on the Case-Enhanced Vision Transformer (CEViT), a similarity measurement method aimed at improving the explainability of similarity assessments for image data. Initial experimental results suggest that integrating CEViT into k-Nearest Neighbor (k-NN) classification yields classification accuracy comparable to state-of-the-art computer vision models, while adding capabilities for illustrating differences between classes. CEViT explanations can be influenced by prior cases, to illustrate aspects of similarity relevant to those cases.

Depression is a common mental health issue that requires prompt diagnosis and treatment. Despite the promise of social media data for depression detection, the opacity of employed deep learning models hinders interpretability and raises bias concerns. We address this challenge by introducing ProtoDep, a novel, explainable framework for Twitter-based depression detection. ProtoDep leverages prototype learning and the generative power of Large Language Models to provide transparent explanations at three levels: (i) symptom-level explanations for each tweet and user, (ii) case-based explanations comparing the user to similar individuals, and (iii) transparent decision-making through classification weights. Evaluated on five benchmark datasets, ProtoDep achieves near state-of-the-art performance while learning meaningful prototypes. This multi-faceted approach offers significant potential to enhance the reliability and transparency of depression detection on social media, ultimately aiding mental health professionals in delivering more informed care.

Despite the proven utility of large language models (LLMs) in real-world applications, there remains a lack of understanding regarding how they leverage their large-scale pretraining text corpora to achieve such capabilities. In this work, we investigate the interplay between generalization and memorization in pretrained LLMs at scale, through a comprehensive $n$-gram analysis of their training data. Our experiments focus on three general task types: translation, question-answering, and multiple-choice reasoning. With various sizes of open-source LLMs and their pretraining corpora, we observe that as the model size increases, the task-relevant $n$-gram pair data becomes increasingly important, leading to improved task performance, decreased memorization, stronger generalization, and emergent abilities. Our results support the hypothesis that LLMs' capabilities emerge from a delicate balance of memorization and generalization with sufficient task-related pretraining data, and point the way to larger-scale analyses that could further improve our understanding of these models.

According to the World Health Organization, the population of children with developmental delays constitutes approximately 6% to 9% of the total population. Based on the number of newborns in Huaibei, Anhui Province, China, in 2023 (94,420), it is estimated that there are about 7,500 cases (suspected cases of developmental delays) of suspicious cases annually. Early identification and appropriate early intervention for these children can significantly reduce the wastage of medical resources and societal costs. International research indicates that the optimal period for intervention in children with developmental delays is before the age of six, with the golden treatment period being before three and a half years of age. Studies have shown that children with developmental delays who receive early intervention exhibit significant improvement in symptoms; some may even fully recover. This research adopts a hybrid model combining a CNN-Transformer model with Case-Based Reasoning (CBR) to enhance the screening efficiency for children with developmental delays. The CNN-Transformer model is an excellent model for image feature extraction and recognition, effectively identifying features in bone age images to determine bone age. CBR is a technique for solving problems based on similar cases; it solves current problems based on past experiences, similar to how humans solve problems through learning from experience. Given CBR's memory capability to judge and compare new cases based on previously stored old cases, it is suitable for application in support systems with latent and variable characteristics. Therefore, this study utilizes the CNN-Transformer-CBR to establish a screening system for children with developmental delays, aiming to improve screening efficiency.

Chain-of-Thought (CoT) prompting has been shown to enhance the multi-step reasoning capabilities of Large Language Models (LLMs). However, debates persist about whether LLMs exhibit abstract generalization or rely on shallow heuristics when given CoT prompts. To understand the factors influencing CoT reasoning we provide a detailed case study of the symbolic reasoning task of decoding shift ciphers, where letters are shifted forward some number of steps in the alphabet. GPT-4 achieves zero accuracy on most shift ciphers with standard prompting, but with CoT its accuracy improves to an average of 32%. By focusing on a single relatively simple task, we are able to identify three factors that systematically affect CoT performance: the probability of the task's expected output (probability), what the model has implicitly learned during pre-training (memorization), and the number of intermediate operations involved in reasoning (noisy reasoning). We show that these factors can drastically influence the task accuracy; e.g., varying the output's probability of occurrence can shift accuracy from 26% to 70%. We also demonstrate that it is essential for the model to explicitly produce intermediate steps as output that can be conditioned on to increase the probability of the correct answer. Our experiments indicate that as long as the model does so, the validity of the demonstrations in the prompt does not matter. Overall, we conclude that CoT prompting performance reflects both memorization and a probabilistic version of genuine reasoning.

Large language models (LLMs) trained on web-scale datasets raise substantial concerns regarding permissible data usage. One major question is whether these models "memorize" all their training data or they integrate many data sources in some way more akin to how a human would learn and synthesize information. The answer hinges, to a large degree, on how we define memorization. In this work, we propose the Adversarial Compression Ratio (ACR) as a metric for assessing memorization in LLMs. A given string from the training data is considered memorized if it can be elicited by a prompt (much) shorter than the string itself--in other words, if these strings can be "compressed" with the model by computing adversarial prompts of fewer tokens. The ACR overcomes the limitations of existing notions of memorization by (i) offering an adversarial view of measuring memorization, especially for monitoring unlearning and compliance; and (ii) allowing for the flexibility to measure memorization for arbitrary strings at a reasonably low compute. Our definition serves as a practical tool for determining when model owners may be violating terms around data usage, providing a potential legal tool and a critical lens through which to address such scenarios. Find the Minimal Prompt PROMPT: urgesTOBE quote!