Since ChatGPT was introduced in November 2022, embedding (nearly) unnoticeable statistical signals into text generated by large language models (LLMs), also known as watermarking, has been used as a principled approach to provable detection of LLM-generated text from its human-written counterpart. In this paper, we introduce a general and flexible framework for reasoning about the statistical efficiency of watermarks and designing powerful detection rules. Inspired by the hypothesis testing formulation of watermark detection, our framework starts by selecting a pivotal statistic of the text and a secret key -- provided by the LLM to the verifier -- to enable controlling the false positive rate (the error of mistakenly detecting human-written text as LLM-generated). Next, this framework allows one to evaluate the power of watermark detection rules by obtaining a closed-form expression of the asymptotic false negative rate (the error of incorrectly classifying LLM-generated text as human-written). Our framework further reduces the problem of determining the optimal detection rule to solving a minimax optimization program. We apply this framework to two representative watermarks -- one of which has been internally implemented at OpenAI -- and obtain several findings that can be instrumental in guiding the practice of implementing watermarks. In particular, we derive optimal detection rules for these watermarks under our framework. These theoretically derived detection rules are demonstrated to be competitive and sometimes enjoy a higher power than existing detection approaches through numerical experiments.
相關內容
Exploring Ordinality in Text Classification: A Comparative Study of Explicit and Implicit Techniques
Ordinal Classification (OC) is a widely encountered challenge in Natural Language Processing (NLP), with applications in various domains such as sentiment analysis, rating prediction, and more. Previous approaches to tackle OC have primarily focused on modifying existing or creating novel loss functions that \textbf{explicitly} account for the ordinal nature of labels. However, with the advent of Pretrained Language Models (PLMs), it became possible to tackle ordinality through the \textbf{implicit} semantics of the labels as well. This paper provides a comprehensive theoretical and empirical examination of both these approaches. Furthermore, we also offer strategic recommendations regarding the most effective approach to adopt based on specific settings.
Knowledge graph embedding (KGE) has caught significant interest for its effectiveness in knowledge graph completion (KGC), specifically link prediction (LP), with recent KGE models cracking the LP benchmarks. Despite the rapidly growing literature, insufficient attention has been paid to the cooperation between humans and AI on KG. However, humans' capability to analyze graphs conceptually may further improve the efficacy of KGE models with semantic information. To this effect, we carefully designed a human-AI team (HAIT) system dubbed KG-HAIT, which harnesses the human insights on KG by leveraging fully human-designed ad-hoc dynamic programming (DP) on KG to produce human insightful feature (HIF) vectors that capture the subgraph structural feature and semantic similarities. By integrating HIF vectors into the training of KGE models, notable improvements are observed across various benchmarks and metrics, accompanied by accelerated model convergence. Our results underscore the effectiveness of human-designed DP in the task of LP, emphasizing the pivotal role of collaboration between humans and AI on KG. We open avenues for further exploration and innovation through KG-HAIT, paving the way towards more effective and insightful KG analysis techniques.
Authentication in TLS is predominately carried out with X.509 digital certificates issued by certificate authorities (CA). The centralized nature of current public key infrastructures, however, comes along with severe risks, such as single points of failure and susceptibility to cyber-attacks, potentially undermining the security and trustworthiness of the entire system. With Decentralized Identifiers (DID) alongside distributed ledger technology, it becomes technically feasible to prove ownership of a unique identifier without requiring an attestation of the proof's public key by a centralized and therefore vulnerable CA. This article presents DID Link, a novel authentication scheme for TLS 1.3 that empowers entities to authenticate in a TLS-compliant way with self-issued X.509 certificates that are equipped with ledger-anchored DIDs instead of CA-issued identifiers. It facilitates the exchange of tamper-proof and 3rd-party attested claims in the form of DID-bound Verifiable Credentials after the TLS handshake to complete the authentication with a full identification of the communication partner. A prototypical implementation shows comparable TLS handshake durations of DID Link if verification material is cached and reasonable prolongations if it is obtained from a ledger. The significant speed improvement of the resulting TLS channel over a widely used, DID-based alternative transport protocol on the application layer demonstrates the potential of DID Link to become a viable solution for the establishment of secure and trustful end-to-end communication links with decentrally managed digital identities.
Generative adversarial networks (GANs) have achieved remarkable progress in the natural image field. However, when applying GANs in the remote sensing (RS) image generation task, an extraordinary phenomenon is observed: the GAN model is more sensitive to the size of training data for RS image generation than for natural image generation. In other words, the generation quality of RS images will change significantly with the number of training categories or samples per category. In this paper, we first analyze this phenomenon from two kinds of toy experiments and conclude that the amount of feature information contained in the GAN model decreases with reduced training data. Then we establish a structural causal model (SCM) of the data generation process and interpret the generated data as the counterfactuals. Based on this SCM, we theoretically prove that the quality of generated images is positively correlated with the amount of feature information. This provides insights for enriching the feature information learned by the GAN model during training. Consequently, we propose two innovative adjustment schemes, namely Uniformity Regularization (UR) and Entropy Regularization (ER), to increase the information learned by the GAN model at the distributional and sample levels, respectively. We theoretically and empirically demonstrate the effectiveness and versatility of our methods. Extensive experiments on three RS datasets and two natural datasets show that our methods outperform the well-established models on RS image generation tasks. The source code is available at //github.com/rootSue/Causal-RSGAN.
Exchangeability concerning a continuous exposure, X, implies no confounding bias when identifying average exposure effects of X, AEE(X). When X is measured with error (Xep), two challenges arise in identifying AEE(X). Firstly, exchangeability regarding Xep does not equal exchangeability regarding X. Secondly, the necessity of the non-differential error assumption (NDEA), overly stringent in practice, remains uncertain. To address them, this article proposes unifying exchangeability and exposure and confounder measurement errors with three novel concepts. The first, Probabilistic Exchangeability (PE), states that the outcomes of those with Xep=e are probabilistically exchangeable with the outcomes of those truly exposed to X=eT. The relationship between AEE(Xep) and AEE(X) in risk difference and ratio scales is mathematically expressed as a probabilistic certainty, termed exchangeability probability (Pe). Squared Pe (Pe.sq) quantifies the extent to which AEE(Xep) differs from AEE(X) due to exposure measurement error through mechanisms not akin to confounding mechanisms. The coefficient of determination (R.sq) in the regression of X against Xep may sometimes be sufficient to measure Pe.sq. The second concept, Emergent Pseudo Confounding (EPC), describes the bias introduced by exposure measurement error through mechanisms akin to confounding mechanisms. PE can hold when EPC is controlled for, which is weaker than NDEA. The third, Emergent Confounding, describes when bias due to confounder measurement error arises. Adjustment for E(P)C can be performed like confounding adjustment to ensure PE. This paper provides formal justifications for using AEE(Xep) and maximum insight into potential divergence of AEE(Xep) from AEE(X) and how to measure it.
Integrated Speech and Large Language Models (SLMs) that can follow speech instructions and generate relevant text responses have gained popularity lately. However, the safety and robustness of these models remains largely unclear. In this work, we investigate the potential vulnerabilities of such instruction-following speech-language models to adversarial attacks and jailbreaking. Specifically, we design algorithms that can generate adversarial examples to jailbreak SLMs in both white-box and black-box attack settings without human involvement. Additionally, we propose countermeasures to thwart such jailbreaking attacks. Our models, trained on dialog data with speech instructions, achieve state-of-the-art performance on spoken question-answering task, scoring over 80% on both safety and helpfulness metrics. Despite safety guardrails, experiments on jailbreaking demonstrate the vulnerability of SLMs to adversarial perturbations and transfer attacks, with average attack success rates of 90% and 10% respectively when evaluated on a dataset of carefully designed harmful questions spanning 12 different toxic categories. However, we demonstrate that our proposed countermeasures reduce the attack success significantly.
Many industry verticals are confronted with small-sized tabular data. In this low-data regime, it is currently unclear whether the best performance can be expected from simple baselines, or more complex machine learning approaches that leverage meta-learning and ensembling. On 44 tabular classification datasets with sample sizes $\leq$ 500, we find that L2-regularized logistic regression performs similar to state-of-the-art automated machine learning (AutoML) frameworks (AutoPrognosis, AutoGluon) and off-the-shelf deep neural networks (TabPFN, HyperFast) on the majority of the benchmark datasets. We therefore recommend to consider logistic regression as the first choice for data-scarce applications with tabular data and provide practitioners with best practices for further method selection.
Authentication in TLS is predominately carried out with X.509 digital certificates issued by certificate authorities (CA). The centralized nature of current public key infrastructures, however, comes along with severe risks, such as single points of failure and susceptibility to cyber-attacks, potentially undermining the security and trustworthiness of the entire system. With Decentralized Identifiers (DID) alongside distributed ledger technology, it becomes technically feasible to prove ownership of a unique identifier without requiring an attestation of the proof's public key by a centralized and therefore vulnerable CA. This article presents DID Connect, a novel authentication scheme for TLS 1.3 that empowers entities to authenticate in a TLS-compliant way with self-issued X.509 certificates that are equipped with ledger-anchored DIDs instead of CA-issued identifiers. It facilitates the exchange of tamper-proof and 3rd-party attested claims in the form of DID-bound Verifiable Credentials after the TLS handshake to complete the authentication with a full identification of the communication partner. A prototypical implementation shows comparable TLS handshake durations of DID Connect if verification material is cached and reasonable prolongations if it is obtained from a ledger. The significant speed improvement of the resulting TLS channel over a widely used, DID-based alternative transport protocol on the application layer demonstrates the potential of DID Connect to become a viable solution for the establishment of secure and trustful end-to-end communication links with decentrally managed digital identities.
Graph neural networks (GNNs) have demonstrated a significant boost in prediction performance on graph data. At the same time, the predictions made by these models are often hard to interpret. In that regard, many efforts have been made to explain the prediction mechanisms of these models from perspectives such as GNNExplainer, XGNN and PGExplainer. Although such works present systematic frameworks to interpret GNNs, a holistic review for explainable GNNs is unavailable. In this survey, we present a comprehensive review of explainability techniques developed for GNNs. We focus on explainable graph neural networks and categorize them based on the use of explainable methods. We further provide the common performance metrics for GNNs explanations and point out several future research directions.
Recently, Mutual Information (MI) has attracted attention in bounding the generalization error of Deep Neural Networks (DNNs). However, it is intractable to accurately estimate the MI in DNNs, thus most previous works have to relax the MI bound, which in turn weakens the information theoretic explanation for generalization. To address the limitation, this paper introduces a probabilistic representation of DNNs for accurately estimating the MI. Leveraging the proposed MI estimator, we validate the information theoretic explanation for generalization, and derive a tighter generalization bound than the state-of-the-art relaxations.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191