论文检索

Social Aspects · Accountability, Transparency and Interpretability

Kiho Park, Yo Joong Choe, Yibo Jiang, Victor Veitch

The linear representation hypothesis is the informal idea that semantic concepts are encoded as linear directions in the representation spaces of large language models (LLMs). Previous work has shown how to make this notion precise for representing binary concepts that have natural contrasts (e.g., {male, female}) as _directions_ in representation space. However, many natural concepts do not have natural contrasts (e.g., whether the output is about an animal). In this work, we show how to extend the formalization of the linear representation hypothesis to represent features (e.g., is_animal) as _vectors_. This allows us to immediately formalize the representation of categorical concepts as polytopes in the representation space. Further, we use the formalization to prove a relationship between the hierarchical structure of concepts and the geometry of their representations. We validate these theoretical results on the Gemma and LLaMA-3 large language models, estimating representations for 900+ hierarchically related concepts using data from WordNet.

Social Aspects · Everything Else

Yue Jiang, Haokun Lin, Yang Bai, Bo Peng, Zhili Liu, Yueming Lyu, Yong Yang, Xingzheng, Jing Dong

Recent studies have discovered that widely used text-to-image diffusion models can replicate training samples during image generation, a phenomenon known as memorization. Existing detection methods primarily focus on identifying memorized prompts. However, in real-world scenarios, image owners may need to verify whether their proprietary or personal images have been memorized by the model, even in the absence of paired prompts or related metadata. We refer to this challenge as image-level memorization detection, where current methods relying on original prompts fall short. In this work, we uncover two characteristics of memorized images after perturbing the inference procedure: lower similarity of the original images and larger magnitudes of TCNP.Building on these insights, we propose Inversion-based Inference Perturbation (IIP), a new framework for image-level memorization detection. Our approach uses unconditional DDIM inversion to derive latent codes that contain core semantic information of original images and optimizes random prompt embeddings to introduce effective perturbation. Memorized images exhibit distinct characteristics within the proposed pipeline, providing a robust basis for detection. To support this task, we construct a comprehensive setup for the image-level memorization detection, carefully curating datasets to simulate realistic memorization scenarios. Using this setup, we evaluate our IIP framework across three different memorization settings, demonstrating its state-of-the-art performance in identifying memorized images in various settings, even in the presence of data augmentation attacks.

Social Aspects · Accountability, Transparency and Interpretability

Christopher Musco, R. Teal Witter

Originally introduced in game theory, Shapley values have emerged as a central tool in explainable machine learning, where they are used to attribute model predictions to specific input features. However, computing Shapley values exactly is expensive: for a model with $n$ features, $O(2^n)$ model evaluations are necessary. To address this issue, approximation algorithms are widely used. One of the most popular is the Kernel SHAP algorithm, which is model agnostic and remarkably effective in practice. However, to the best of our knowledge, Kernel SHAP has no strong non-asymptotic complexity guarantees. We address this issue by introducing *Leverage SHAP*, a light-weight modification of Kernel SHAP that provides provably accurate Shapley value estimates with just $O(n\log n)$ model evaluations. Our approach takes advantage of a connection between Shapley value estimation and agnostic active learning by employing *leverage score sampling*, a powerful regression tool. Beyond theoretical guarantees, we show that Leverage SHAP consistently outperforms even the highly optimized implementation of Kernel SHAP available in the ubiquitous SHAP library [Lundberg \& Lee, 2017].

Social Aspects · Trustworthy Machine Learning

Juyeon Heo, Miao Xiong, Christina Heinze-Deml, Jaya Narain

Large language models (LLMs) could be valuable personal AI agents across various domains, provided they can precisely follow user instructions. However, recent studies have shown significant limitations in LLMs' instruction-following capabilities, raising concerns about their reliability in high-stakes applications. Accurately estimating LLMs' uncertainty in adhering to instructions is critical to mitigating deployment risks. We present, to our knowledge, the first systematic evaluation of the uncertainty estimation abilities of LLMs in the context of instruction-following. Our study identifies key challenges with existing instruction-following benchmarks, where multiple factors are entangled with uncertainty stems from instruction-following, complicating the isolation and comparison across methods and models.To address these issues, we introduce a controlled evaluation setup with two benchmark versions of data, enabling a comprehensive comparison of uncertainty estimation methods under various conditions.Our findings show that existing uncertainty methods struggle, particularly when models make subtle errors in instruction following. While internal model states provide some improvement, they remain inadequate in more complex scenarios. The insights from our controlled evaluation setups provide a crucial understanding of LLMs' limitations and potential for uncertainty estimation in instruction-following tasks, paving the way for more trustworthy AI agents.

Social Aspects · Accountability, Transparency and Interpretability

Shayne Longpre, Nikhil Singh, Manuel Cherep, Kushagra Tiwary, Joanna Materzynska, William Brannon, Robert Mahari, Naana Obeng-Marnu, Manan Dey, Mohammed Hamdy 等

Progress in AI is driven largely by the scale and quality of training data. Despite this, there is a deficit of empirical analysis examining the attributes of well-established datasets beyond text. In this work we conduct the largest and first-of-its-kind longitudinal audit across modalities --- popular text, speech, and video datasets --- from their detailed sourcing trends and use restrictions to their geographical and linguistic representation. Our manual analysis covers nearly 4000 public datasets between 1990-2024, spanning 608 languages, 798 sources, 659 organizations, and 67 countries. We find that multimodal machine learning applications have overwhelmingly turned to web-crawled, synthetic, and social media platforms, such as YouTube, for their training sets, eclipsing all other sources since 2019. Secondly, tracing the chain of dataset derivations we find that while less than 33% of datasets are restrictively licensed, over 80% of the source content in widely-used text, speech, and video datasets, carry non-commercial restrictions. Finally, counter to the rising number of languages and geographies represented in public AI training datasets, our audit demonstrates measures of relative geographical and multilingual representation have failed to significantly improve their coverage since 2013. We believe the breadth of our audit enables us to empirically examine trends in data sourcing, restrictions, and Western-centricity at an ecosystem-level, and that visibility into these questions are essential to progress in responsible AI. As a contribution to ongoing improvements in dataset transparency and responsible use, we release our entire multimodal audit, allowing practitioners to trace data provenance across text, speech, and video.

Social Aspects · Fairness, Equity, Justice and Safety

Maksym Andriushchenko, Alexandra Souly, Mateusz Dziemian, Derek Duenas, Maxwell Lin, Justin Wang, Dan Hendrycks, Andy Zou, Zico Kolter, Matt Fredrikson 等

The robustness of LLMs to jailbreak attacks, where users design prompts to circumvent safety measures and misuse model capabilities, has been studied primarily for LLMs acting as simple chatbots. Meanwhile, LLM agents---which use external tools and can execute multi-stage tasks---may pose a greater risk if misused, but their robustness remains underexplored. To facilitate research on LLM agent misuse, we propose a new benchmark called AgentHarm. The benchmark includes a diverse set of 110 explicitly malicious agent tasks (440 with augmentations), covering 11 harm categories including fraud, cybercrime, and harassment. In addition to measuring whether models refuse harmful agentic requests, scoring well on AgentHarm requires jailbroken agents to maintain their capabilities following an attack to complete a multi-step task. We evaluate a range of leading LLMs, and find (1) leading LLMs are surprisingly complaint with malicious agent requests without jailbreaking, (2) simple universal jailbreak strings can be adapted to effectively jailbreak agents, and (3) these jailbreaks enable coherent and malicious multi-step agent behavior and retain model capabilities. To enable simple and reliable evaluation of attacks and defenses for LLM-based agents, we publicly release AgentHarm at https://huggingface.co/datasets/ai-safety-institute/AgentHarm.

Social Aspects · Fairness, Equity, Justice and Safety

Tinghao Xie, Xiangyu Qi, Yi Zeng, Yangsibo Huang, Udari Sehwag, Kaixuan Huang, Luxi He, Boyi Wei, Dacheng Li, Ying Sheng 等

Evaluating aligned large language models' (LLMs) ability to recognize and reject unsafe user requests is crucial for safe, policy-compliant deployments. Existing evaluation efforts, however, face three limitations that we address with **SORRY-Bench**, our proposed benchmark. **First**, existing methods often use coarse-grained taxonomies of unsafe topics, and are over-representing some fine-grained topics. For example, among the ten existing datasets that we evaluated, tests for refusals of self-harm instructions are over 3x less represented than tests for fraudulent activities. SORRY-Bench improves on this by using a fine-grained taxonomy of 44 potentially unsafe topics, and 440 class-balanced unsafe instructions, compiled through human-in-the-loop methods. **Second**, evaluations often overlook the linguistic formatting of prompts, like different languages, dialects, and more --- which are only implicitly considered in many evaluations. We supplement SORRY-bench with 20 diverse linguistic augmentations to systematically examine these effects. **Third**, existing evaluations rely on large LLMs (e.g., GPT-4) for evaluation, which can be computationally expensive. We investigate design choices for creating a fast, accurate automated safety evaluator. By collecting 7K+ human annotations and conducting a meta-evaluation of diverse LLM-as-a-judge designs, we show that fine-tuned 7B LLMs can achieve accuracy comparable to GPT-4 scale LLMs, with lower computational cost. Putting these together, we evaluate over 50 proprietary and open-weight LLMs on SORRY-Bench, analyzing their distinctive safety refusal behaviors. We hope our effort provides a building block for systematic evaluations of LLMs' safety refusal capabilities, in a balanced, granular, and efficient manner. Benchmark demo, data, code, and models are available through [https://sorry-bench.github.io](https://sorry-bench.github.io).

Kaicheng Yu, Zhuang Shao, Siyuan Qi, Dongfang Liu

The tutorial "Large Vision-Language Model in the Society" aims to provide a comprehensive overview of state-of-the-art techniques and applications of large vision-language models (LVLMs), which integrate visual and textual data to transform multimedia research and applications. LVLMs are poised to revolutionize domains such as content creation, social media analysis, education, healthcare, and entertainment by enabling sophisticated content analysis, retrieval, and generation. This tutorial will cover the fundamentals of vision-language integration, state-of-the-art models, training techniques, applications, ethical considerations, and future directions. It is designed to be educational and instructive, providing an in-depth introduction rather than a cursory survey. Attendees will gain practical skills, and insights into the latest research, and engage in interactive sessions to reinforce learning. By addressing both technical and societal aspects, the tutorial will significantly benefit the multimedia community, driving innovation and progress in the field.

Federico Espositi, Andrea Bonarini

We are investigating interaction with entities showing features different from humans', to understand how they can be embodied as avatars and perceived as living, social beings. To push this investigation to its limit, we have designed as an avatar an interactive space (the Room), that challenges both the anthropomorphic structure, and most of the social interaction mechanisms we are used to. We introduce a first framework for the Room design, addressing challenges related to its body, perception, and interaction process. We present a pilot implementation of some of the aspects of the framework as an interactive installation, namely a real-time, two-player, VR experience, featuring the Room avatar, with a focus on haptic feedback as the main means of perception for the subject embodying the Room. By radically challenging anthropomorphism, we seek to investigate the most basic aspects of embodiment and social cognition.

Yupeng Li 0001, Haorui He, Jin Bai, Dacheng Wen

The prevalence of fake news across various online sources has had a significant influence on the public. Existing Chinese fake news detection datasets are limited to news sourced solely from Weibo. However, fake news originating from multiple sources exhibits diversity in various aspects, including its content and social context. Methods trained on purely one single news source can hardly be applicable to real-world scenarios. Our pilot experiment demonstrates that the F1 score of the state-of-the-art method that learns from a large Chinese fake news detection dataset, Weibo-21, drops significantly from 0.943 to 0.470 when the test data is changed to multi-source news data, failing to identify more than one-third of the multi-source fake news. To address this limitation, we constructed the first multi-source benchmark dataset for Chinese fake news detection, termed MCFEND, which is composed of news we collected from diverse sources such as social platforms, messaging apps, and traditional online news outlets. Notably, such news has been fact-checked by 14 authoritative fact-checking agencies worldwide. In addition, various existing Chinese fake news detection methods are thoroughly evaluated on our proposed dataset in cross-source, multi-source, and unseen source ways. MCFEND, as a benchmark dataset, aims to advance Chinese fake news detection approaches in real-world scenarios.

Yuxia Wu, Yuan Fang 0001, Lizi Liao

Dynamic graph modeling is crucial for understanding complex structures in web graphs, spanning applications in social networks, recommender systems, and more. Most existing methods primarily emphasize structural dependencies and their temporal changes. However, these approaches often overlook detailed temporal aspects or struggle with long-term dependencies. Furthermore, many solutions overly complicate the process by emphasizing intricate module designs to capture dynamic evolutions. In this work, we harness the strength of the Transformer's self-attention mechanism, known for adeptly handling long-range dependencies in sequence modeling. Our approach offers a simple Transformer model, called SimpleDyG, tailored for dynamic graph modeling without complex modifications. We re-conceptualize dynamic graphs as a sequence modeling challenge and introduce a novel temporal alignment technique. This technique not only captures the inherent temporal evolution patterns within dynamic graphs but also streamlines the modeling process of their evolution. To evaluate the efficacy of SimpleDyG, we conduct extensive experiments on four real-world datasets from various domains. The results demonstrate the competitive performance of SimpleDyG in comparison to a series of state-of-the-art approaches despite its simple design.

Alexander Rodríguez, Bijaya Adhikari, Ajitesh Srivastava, Sen Pei, Marie-Laure Charpignon, Kai Wang 0040, Serina Chang, Anil Vullikanti, B. Aditya Prakash

While the worst of COVID-19 pandemic has most likely passed us, an occurrence of equally devastating global pandemic or regional epidemic cannot be ruled out in future. H1N1, Zika, SARS, MERS, and Ebola outbreaks over the past few decades have sharply illustrated our enormous vulnerability to emerging infectious diseases. While the data mining research community has demonstrated increased interest in epidemiological applications, much is still left to be desired. For example, there is an urgent need to develop sound theoretical principles and transformative computational approaches that will allow us to address the escalating threat of current and future pandemics. Data mining and knowledge discovery have an important role to play in this regard. Different aspects of infectious disease modeling, analysis, and control have traditionally been studied within the confines of individual disciplines, such as mathematical epidemiology and public health, and data mining and machine learning. Coupled with increasing data generation across multiple domains/sources (e.g., wastewater surveillance, electronic medical records, and social media), there is a clear need for analyzing them to inform public health policies and outcomes timely. Recent advances in disease surveillance and forecasting, and initiatives such as the CDC Flu Challenge, CDC COVID-19 Forecasting Hub etc., have brought these disciplines closer together. On the one hand, public health practitioners seek to use novel datasets, such as Safegraph, Unacast, and Google mobility data, and techniques like Graph Neural Networks. On the other hand, researchers from data mining and machine learning develop novel tools for solving many fundamental problems in the public health policy planning and decision-making process, leveraging novel datasets (e.g., COVID-19 behavioral health surveys, contact tracing trees, and satellite images of urban streets) and combining them with more traditional time series information (e.g., surveillance, hospitalization, and death records). We believe the next stage of advances will result from closer collaborations between these two groups, which is the main objective of epiDAMIK.

Jia Li 0009, Xiangguo Sun, Yuhan Li 0001, Zhixun Li, Hong Cheng 0001, Jeffrey Xu Yu

Graph plays a significant role in representing and analyzing complex relationships in real-world applications such as citation networks, social networks, and biological data. Graph intelligence is rapidly becoming a crucial aspect of understanding and exploiting the intricate interconnections within graph data. Recently, large language models (LLMs) and prompt learning techniques have pushed graph intelligence forward, outperforming traditional Graph Neural Network (GNN) pre-training methods and setting new benchmarks for performance. In this tutorial, we begin by offering a comprehensive review and analysis of existing methods that integrate LLMs with graphs. We introduce existing works based on a novel taxonomy that classifies them into three distinct categories according to the roles of LLMs in graph tasks: as enhancers, predictors, or alignment components. Secondly, we introduce a new learning method that utilizes prompting on graphs, offering substantial potential to enhance graph transfer capabilities across diverse tasks and domains. We discuss existing works on graph prompting within a unified framework and introduce our developed tool for executing a variety of graph prompting tasks. Additionally, we discuss the applications of combining Graphs, LLMs, and prompt learning across various tasks, such as urban computing, recommendation systems, and anomaly detection. This lecture-style tutorial is an extension of our original work published in IJCAI 2024[44] and arXiv[77] with the invitation of KDD24.

Zhangtao Cheng, Jienan Zhang, Xovee Xu, Goce Trajcevski, Ting Zhong, Fan Zhou 0002

Accurately predicting the popularity of multimodal user-generated content (UGC) is fundamental for many real-world applications such as online advertising and recommendation. Existing approaches generally focus on limited contextual information within individual UGCs, yet overlook the potential benefit of exploiting meaningful knowledge in relevant UGCs. In this work, we propose RAGTrans, an aspect-aware retrieval-augmented multi-modal hypergraph transformer that retrieves pertinent knowledge from a multi-modal memory bank and enhances UGC representations via neighborhood knowledge aggregation on multi-model hypergraphs. In particular, we initially retrieve relevant multimedia instances from a large corpus of UGCs via the aspect information and construct a knowledge-enhanced hypergraph based on retrieved relevant instances. This allows capturing meaningful contextual information across the data. We then design a novel bootstrapping hypergraph transformer on multimodal hypergraphs to strengthen UGC representations across modalities via customizing a propagation algorithm to effectively diffuse information across nodes and edges. Additionally, we propose a user-aware attention-based fusion module to comprise the enriched UGC representations for popularity prediction. Extensive experiments on real-world social media datasets demonstrate that RAGTrans outperforms state-of-the-art popularity prediction models across settings.

Wei Ju, Siyu Yi, Yifan Wang, Qingqing Long, Junyu Luo, Zhiping Xiao, Ming Zhang

Graph-structured data, prevalent in domains ranging from social networks to biochemical analysis, serve as the foundation for diverse real-world systems. While graph neural networks demonstrate proficiency in modeling this type of data, their success is often reliant on significant amounts of labeled data, posing a challenge in practical scenarios with limited annotation resources. To tackle this problem, tremendous efforts have been devoted to enhancing graph machine learning performance under low-resource settings by exploring various approaches to minimal supervision. In this paper, we introduce a novel concept of Data-Efficient Graph Learning (DEGL) as a research frontier, and present the first survey that summarizes the current progress of DEGL. We initiate by highlighting the challenges inherent in training models with large labeled data, paving the way for our exploration into DEGL. Next, we systematically review recent advances on this topic from several key aspects, including self-supervised graph learning, semi-supervised graph learning, and few-shot graph learning. Also, we state promising directions for future research, contributing to the evolution of graph machine learning.

Hadi Hosseini, Shraddha Pathak

Matching markets consist of two disjoint sets of agents, where each agent has a preference list over agents on the other side. The primary objective is to find a stable matching between the agents such that no unmatched pair of agents prefer each other to their matched partners. The incompatibility between stability and strategy-proofness in this domain gives rise to a variety of strategic behavior of agents, which in turn may influence the resulting matching. In this paper, we discuss fundamental properties of stable matchings, review essential structural observations, survey key results in manipulation algorithms and their game-theoretical aspects, and more importantly, highlight a series of open research questions.

Siyue Ren, Zhiyao Cui, Ruiqi Song, Zhen Wang, Shuyue Hu

Social norms play a crucial role in guiding agents towards understanding and adhering to standards of behavior, thus reducing social conflicts within multi-agent systems (MASs). However, current LLM-based (or generative) MASs lack the capability to be normative. In this paper, we propose a novel architecture, named CRSEC, to empower the emergence of social norms within generative MASs. Our architecture consists of four modules: Creation & Representation, Spreading, Evaluation, and Compliance. This addresses several important aspects of the emergent processes all in one: (i) where social norms come from, (ii) how they are formally represented, (iii) how they spread through agents' communications and observations, (iv) how they are examined with a sanity check and synthesized in the long term, and (v) how they are incorporated into agents' planning and actions. Our experiments deployed in the Smallville sandbox game environment demonstrate the capability of our architecture to establish social norms and reduce social conflicts within generative MASs. The positive outcomes of our human evaluation, conducted with 30 evaluators, further affirm the effectiveness of our approach. Our project can be accessed via the following link: https://github.com/sxswz213/CRSEC.

Jinjin Li, Bin Liu, Chengyan Liu, Hongli Zhang

Urban migration is a significant aspect of a city's economy. The exploration of the underlying determinants of housing purchases among current residents contributes to the study of future trends in urban migration, enabling governments to formulate appropriate policies to guide future economic growth. This article employs a factor model to analyze data on residents' rentals, first-time home purchases, and subsequent housing upgrades. We decompose the factors influencing housing purchases into common drivers and specific drivers. Our hypothesis is that common drivers reflect universal social patterns, while personalized drivers represent stochastic elements. We construct a correlation matrix capturing the inter-resident relationships based on the common drivers of housing purchases. We then propose a graph neural network based on the correlation matrix to model housing predictions as a node classification problem. Our model addresses two critical questions. Firstly, we aim to identify which part of rental residents will engage in first-time home purchases in the future. Secondly, we seek to determine which group of residents, having completed rental and first-time home purchases, will opt for a second home purchase. The results of our testing on real-world datasets demonstrate that based solely on rental and home purchase records, we can achieve a sensitivity for housing predictions exceeding 80%.

Théo Delemazure, Łukasz Janeczko, Andrzej Kaczmarczyk, Stanisław Szufa

We study committee elections from a perspective of finding the most conflicting candidates, that is, candidates that imply the largest amount of conflict, as per voter preferences. By proposing basic axioms to capture this objective, we show that none of the prominent multiwinner voting rules meet them. Consequently, we design committee voting rules compliant with our desiderata, introducing conflictual voting rules. A subsequent deepened analysis sheds more light on how they operate. Our investigation identifies various aspects of conflict, for which we come up with relevant axioms and quantitative measures, which may be of independent interest. We support our theoretical study with experiments on both real-life and synthetic data.

Gennaro Auricchio, Zihe Wang, Jie Zhang

In this paper, we investigate the Mechanism Design aspects of the m-Capacitated Facility Location Problem (m-CFLP) on a line. We focus on two frameworks. In the first framework, the number of facilities is arbitrary, all facilities have the same capacity, and the number of agents is equal to the total capacity of all facilities. In the second framework, we aim to place two facilities, each with a capacity of at least half of the total agents. For both of these frameworks, we propose truthful mechanisms with bounded approximation ratios with respect to the Social Cost (SC) and the Maximum Cost (MC). When m>2, the result sharply contrasts with the impossibility results known for the classic m-Facility Location Problem, where capacity constraints are not considered. Furthermore, all our mechanisms are optimal with respect to the MC and optimal or nearly optimal with respect to the SC among anonymous mechanisms. For both frameworks, we provide a lower bound on the approximation ratio that any truthful and deterministic mechanism can achieve with respect to the SC and MC.