Large language models (LLMs) with powerful generalization ability has been widely used in many domains. A systematic and reliable evaluation of LLMs is a crucial step in their development and applications, especially for specific professional fields. In the urban domain, there have been some early explorations about the usability of LLMs, but a systematic and scalable evaluation benchmark is still lacking. The challenge in constructing a systematic evaluation benchmark for the urban domain lies in the diversity of data and scenarios, as well as the complex and dynamic nature of cities. In this paper, we propose CityBench, an interactive simulator based evaluation platform, as the first systematic evaluation benchmark for the capability of LLMs for urban domain. First, we build CitySim to integrate the multi-source data and simulate fine-grained urban dynamics. Based on CitySim, we design 7 tasks in 2 categories of perception-understanding and decision-making group to evaluate the capability of LLMs as city-scale world model for urban domain. Due to the flexibility and ease-of-use of CitySim, our evaluation platform CityBench can be easily extended to any city in the world. We evaluate 13 well-known LLMs including open source LLMs and commercial LLMs in 13 cities around the world. Extensive experiments demonstrate the scalability and effectiveness of proposed CityBench and shed lights for the future development of LLMs in urban domain. The dataset, benchmark and source codes are openly accessible to the research community via https://github.com/tsinghua-fib-lab/CityBench

Human Activity Recognition is a time-series analysis problem. A popular analysis procedure used by the community assumes an optimal window length to design recognition pipelines. However, in the scenario of smart homes, where activities are of varying duration and frequency, the assumption of a constant sized window does not hold. Additionally, previous works have shown these activities to be made up of building blocks. We focus on identifying these underlying building blocks--structural constructs, with the use of large language models. Identifying these constructs can be beneficial especially in recognizing short-duration and infrequent activities. We also propose the development of an activity recognition procedure that uses these building blocks to model activities, thus helping the downstream task of activity monitoring in smart homes.

Large language models (LLMs) have transformed the field of natural language processing, but they remain susceptible to jailbreaking attacks that exploit their capabilities to generate unintended and potentially harmful content. Existing token-level jailbreaking techniques, while effective, face scalability and efficiency challenges, especially as models undergo frequent updates and incorporate advanced defensive measures. In this paper, we introduce JailMine, an innovative token-level manipulation approach that addresses these limitations effectively. JailMine employs an automated "mining" process to elicit malicious responses from LLMs by strategically selecting affirmative outputs and iteratively reducing the likelihood of rejection. Through rigorous testing across multiple well-known LLMs and datasets, we demonstrate JailMine's effectiveness and efficiency, achieving a significant average reduction of 86% in time consumed while maintaining high success rates averaging 95%, even in the face of evolving defensive strategies. Our work contributes to the ongoing effort to assess and mitigate the vulnerability of LLMs to jailbreaking attacks, underscoring the importance of continued vigilance and proactive measures to enhance the security and reliability of these powerful language models.

Unsupervised graph alignment finds the one-to-one node correspondence between a pair of attributed graphs by only exploiting graph structure and node features. One category of existing works first computes the node representation and then matches nodes with close embeddings, which is intuitive but lacks a clear objective tailored for graph alignment in the unsupervised setting. The other category reduces the problem to optimal transport (OT) via Gromov-Wasserstein (GW) learning with a well-defined objective but leaves a large room for exploring the design of transport cost. We propose a principled approach to combine their advantages motivated by theoretical analysis of model expressiveness. By noticing the limitation of discriminative power in separating matched and unmatched node pairs, we improve the cost design of GW learning with feature transformation, which enables feature interaction across dimensions. Besides, we propose a simple yet effective embedding-based heuristic inspired by the Weisfeiler-Lehman test and add its prior knowledge to OT for more expressiveness when handling non-Euclidean data. Moreover, we are the first to guarantee the one-to-one matching constraint by reducing the problem to maximum weight matching. The algorithm design effectively combines our OT and embedding-based predictions via stacking, an ensemble learning strategy. We propose a model framework named \texttt{CombAlign} integrating all the above modules to refine node alignment progressively. Through extensive experiments, we demonstrate significant improvements in alignment accuracy compared to state-of-the-art approaches and validate the effectiveness of the proposed modules.

Stance detection seeks to identify the viewpoints of individuals either in favor or against a given target or a controversial topic. Current advanced neural models for stance detection typically employ fully parametric softmax classifiers. However, these methods suffer from several limitations, including lack of explainability, insensitivity to the latent data structure, and unimodality, which greatly restrict their performance and applications. To address these challenges, we present a novel collaborative stance detection framework called (CoSD) which leverages contrastive heterogeneous topic graph learning to learn topic-aware semantics and collaborative signals among texts, topics, and stance labels for enhancing stance detection. During training, we construct a heterogeneous graph to structurally organize texts and stances through implicit topics via employing latent Dirichlet allocation. We then perform contrastive graph learning to learn heterogeneous node representations, aggregating informative multi-hop collaborative signals via an elaborate Collaboration Propagation Aggregation (CPA) module. During inference, we introduce a hybrid similarity scoring module to enable the comprehensive incorporation of topic-aware semantics and collaborative signals for stance detection. Extensive experiments on two benchmark datasets demonstrate the state-of-the-art detection performance of CoSD, verifying the effectiveness and explainability of our collaborative framework.

In the Retrieval-Augmented Generation (RAG) system, advanced Large Language Models (LLMs) have emerged as effective Query Likelihood Models (QLMs) in an unsupervised way, which re-rank documents based on the probability of generating the query given the content of a document. However, directly prompting LLMs to approximate QLMs inherently is biased, where the estimated distribution might diverge from the actual document-specific distribution. In this study, we introduce a novel framework, $\mathrm{UR^3}$, which leverages Bayesian decision theory to both quantify and mitigate this estimation bias. Specifically, $\mathrm{UR^3}$ reformulates the problem as maximizing the probability of document generation, thereby harmonizing the optimization of query and document generation probabilities under a unified risk minimization objective. Our empirical results indicate that $\mathrm{UR^3}$ significantly enhances re-ranking, particularly in improving the Top-1 accuracy. It benefits the QA tasks by achieving higher accuracy with fewer input documents.

Large Language Models (LLMs) tend to be unreliable in the factuality of their answers. To address this problem, NLP researchers have proposed a range of techniques to estimate LLM's confidence over facts. However, due to the lack of a systematic comparison, it is not clear how the different methods compare to one another. To fill this gap, we present a survey and empirical comparison of estimators of factual confidence. We define an experimental framework allowing for fair comparison, covering both fact-verification and question answering. Our experiments across a series of LLMs indicate that trained hidden-state probes provide the most reliable confidence estimates, albeit at the expense of requiring access to weights and training data. We also conduct a deeper assessment of factual confidence by measuring the consistency of model behavior under meaning-preserving variations in the input. We find that the confidence of LLMs is often unstable across semantically equivalent inputs, suggesting that there is much room for improvement of the stability of models' parametric knowledge. Our code is available at (https://github.com/amazon-science/factual-confidence-of-llms).

Graph neural networks (GNNs) have achieved tremendous success, but recent studies have shown that GNNs are vulnerable to adversarial attacks, which significantly hinders their use in safety-critical scenarios. Therefore, the design of robust GNNs has attracted increasing attention. However, existing research has mainly been conducted via experimental trial and error, and thus far, there remains a lack of a comprehensive understanding of the vulnerability of GNNs. To address this limitation, we systematically investigate the adversarial robustness of GNNs by considering graph data patterns, model-specific factors, and the transferability of adversarial examples. Through extensive experiments, a set of principled guidelines is obtained for improving the adversarial robustness of GNNs, for example: (i) rather than highly regular graphs, the training graph data with diverse structural patterns is crucial for model robustness, which is consistent with the concept of adversarial training; (ii) the large model capacity of GNNs with sufficient training data has a positive effect on model robustness, and only a small percentage of neurons in GNNs are affected by adversarial attacks; (iii) adversarial transfer is not symmetric and the adversarial examples produced by the small-capacity model have stronger adversarial transferability. This work illuminates the vulnerabilities of GNNs and opens many promising avenues for designing robust GNNs.

Productive interactions between diverse users and language technologies require outputs from the latter to be culturally relevant and sensitive. Prior works have evaluated models' knowledge of cultural norms, values, and artifacts, without considering how this knowledge manifests in downstream applications. In this work, we focus on extrinsic evaluation of cultural competence in two text generation tasks, open-ended question answering and story generation. We quantitatively and qualitatively evaluate model outputs when an explicit cue of culture, specifically nationality, is perturbed in the prompts. Although we find that model outputs do vary when varying nationalities and feature culturally relevant words, we also find weak correlations between text similarity of outputs for different countries and the cultural values of these countries. Finally, we discuss important considerations in designing comprehensive evaluation of cultural competence in user-facing tasks.

The advancement of large language models (LLMs) has demonstrated strong capabilities across various applications, including mental health analysis. However, existing studies have focused on predictive performance, leaving the critical issue of fairness underexplored, posing significant risks to vulnerable populations. Despite acknowledging potential biases, previous works have lacked thorough investigations into these biases and their impacts. To address this gap, we systematically evaluate biases across seven social factors (e.g., gender, age, religion) using ten LLMs with different prompting methods on eight diverse mental health datasets. Our results show that GPT-4 achieves the best overall balance in performance and fairness among LLMs, although it still lags behind domain-specific models like MentalRoBERTa in some cases. Additionally, our tailored fairness-aware prompts can effectively mitigate bias in mental health predictions, highlighting the great potential for fair analysis in this field.