Yixuan Zhang

Haonan Li, Yixuan Zhang, Fajri Koto et al.

As the capabilities of large language models (LLMs) continue to advance, evaluating their performance becomes increasingly crucial and challenging. This paper aims to bridge this gap by introducing CMMLU, a comprehensive Chinese benchmark that covers various subjects, including natural science, social sciences, engineering, and humanities. We conduct a thorough evaluation of 18 advanced multilingual- and Chinese-oriented LLMs, assessing their performance across different subjects and settings. The results reveal that most existing LLMs struggle to achieve an average accuracy of 50%, even when provided with in-context examples and chain-of-thought prompts, whereas the random baseline stands at 25%. This highlights significant room for improvement in LLMs. Additionally, we conduct extensive experiments to identify factors impacting the models' performance and propose directions for enhancing LLMs. CMMLU fills the gap in evaluating the knowledge and reasoning capabilities of large language models within the Chinese context.

Qinlin Zhao, Jindong Wang, Yixuan Zhang et al.

Large language models (LLMs) have been widely used as agents to complete different tasks, such as personal assistance or event planning. While most of the work has focused on cooperation and collaboration between agents, little work explores competition, another important mechanism that promotes the development of society and economy. In this paper, we seek to examine the competition dynamics in LLM-based agents. We first propose a general framework for studying the competition between agents. Then, we implement a practical competitive environment using GPT-4 to simulate a virtual town with two types of agents, restaurant agents and customer agents. Specifically, the restaurant agents compete with each other to attract more customers, where competition encourages them to transform, such as cultivating new operating strategies. Simulation experiments reveal several interesting findings at the micro and macro levels, which align well with existing market and sociological theories. We hope that the framework and environment can be a promising testbed to study competition that fosters understanding of society. Code is available at: https://github.com/microsoft/competeai.

Cheng Li, Jindong Wang, Yixuan Zhang et al.

Emotional intelligence significantly impacts our daily behaviors and interactions. Although Large Language Models (LLMs) are increasingly viewed as a stride toward artificial general intelligence, exhibiting impressive performance in numerous tasks, it is still uncertain if LLMs can genuinely grasp psychological emotional stimuli. Understanding and responding to emotional cues gives humans a distinct advantage in problem-solving. In this paper, we take the first step towards exploring the ability of LLMs to understand emotional stimuli. To this end, we first conduct automatic experiments on 45 tasks using various LLMs, including Flan-T5-Large, Vicuna, Llama 2, BLOOM, ChatGPT, and GPT-4. Our tasks span deterministic and generative applications that represent comprehensive evaluation scenarios. Our automatic experiments show that LLMs have a grasp of emotional intelligence, and their performance can be improved with emotional prompts (which we call "EmotionPrompt" that combines the original prompt with emotional stimuli), e.g., 8.00% relative performance improvement in Instruction Induction and 115% in BIG-Bench. In addition to those deterministic tasks that can be automatically evaluated using existing metrics, we conducted a human study with 106 participants to assess the quality of generative tasks using both vanilla and emotional prompts. Our human study results demonstrate that EmotionPrompt significantly boosts the performance of generative tasks (10.9% average improvement in terms of performance, truthfulness, and responsibility metrics). We provide an in-depth discussion regarding why EmotionPrompt works for LLMs and the factors that may influence its performance. We posit that EmotionPrompt heralds a novel avenue for exploring interdisciplinary knowledge for human-LLMs interaction.

Yuan Li, Lichao Sun, Yixuan Zhang

Significant advancements have occurred in the application of Large Language Models (LLMs) for social simulations. Despite this, their abilities to perform teaming in task-oriented social events are underexplored. Such capabilities are crucial if LLMs are to effectively mimic human-like social behaviors and form efficient teams to solve tasks. To bridge this gap, we introduce MetaAgents, a social simulation framework populated with LLM-based agents. MetaAgents facilitates agent engagement in conversations and a series of decision making within social contexts, serving as an appropriate platform for investigating interactions and interpersonal decision-making of agents. In particular, we construct a job fair environment as a case study to scrutinize the team assembly and skill-matching behaviors of LLM-based agents. We take advantage of both quantitative metrics evaluation and qualitative text analysis to assess their teaming abilities at the job fair. Our evaluation demonstrates that LLM-based agents perform competently in making rational decisions to develop efficient teams. However, we also identify limitations that hinder their effectiveness in more complex team assembly tasks. Our work provides valuable insights into the role and evolution of LLMs in task-oriented social simulations.

Yixuan Zhang, Haonan Li

Large language models (LLMs) have showcased remarkable capabilities in understanding and generating language. However, their ability in comprehending ancient languages, particularly ancient Chinese, remains largely unexplored. To bridge this gap, we present ACLUE, an evaluation benchmark designed to assess the capability of language models in comprehending ancient Chinese. ACLUE consists of 15 tasks cover a range of skills, spanning phonetic, lexical, syntactic, semantic, inference and knowledge. Through the evaluation of eight state-of-the-art LLMs, we observed a noticeable disparity in their performance between modern Chinese and ancient Chinese. Among the assessed models, ChatGLM2 demonstrates the most remarkable performance, achieving an average score of 37.4%. We have made our code and data public available.

Zhilin Zhao, Longbing Cao, Yixuan Zhang et al.

Out-of-distribution (OOD) detection is critical for identifying test samples that deviate from in-distribution (ID) data, ensuring network robustness and reliability. This paper presents a flexible framework for OOD knowledge distillation that extracts OOD-sensitive information from a network to develop a binary classifier capable of distinguishing between ID and OOD samples in both scenarios, with and without access to training ID data. To accomplish this, we introduce Confidence Amendment (CA), an innovative methodology that transforms an OOD sample into an ID one while progressively amending prediction confidence derived from the network to enhance OOD sensitivity. This approach enables the simultaneous synthesis of both ID and OOD samples, each accompanied by an adjusted prediction confidence, thereby facilitating the training of a binary classifier sensitive to OOD. Theoretical analysis provides bounds on the generalization error of the binary classifier, demonstrating the pivotal role of confidence amendment in enhancing OOD sensitivity. Extensive experiments spanning various datasets and network architectures confirm the efficacy of the proposed method in detecting OOD samples.

Yifan Sun, Yixuan Zhang, Ali Mosallaei et al.

Graphics Processing Units (GPUs) have been widely used to accelerate artificial intelligence, physics simulation, medical imaging, and information visualization applications. To improve GPU performance, GPU hardware designers need to identify performance issues by inspecting a huge amount of simulator-generated traces. Visualizing the execution traces can reduce the cognitive burden of users and facilitate making sense of behaviors of GPU hardware components. In this paper, we first formalize the process of GPU performance analysis and characterize the design requirements of visualizing execution traces based on a survey study and interviews with GPU hardware designers. We contribute data and task abstraction for GPU performance analysis. Based on our task analysis, we propose Daisen, a framework that supports data collection from GPU simulators and provides visualization of the simulator-generated GPU execution traces. Daisen features a data abstraction and trace format that can record simulator-generated GPU execution traces. Daisen also includes a web-based visualization tool that helps GPU hardware designers examine GPU execution traces, identify performance bottlenecks, and verify performance improvement. Our qualitative evaluation with GPU hardware designers demonstrates that the design of Daisen reflects the typical workflow of GPU hardware designers. Using Daisen, participants were able to effectively identify potential performance bottlenecks and opportunities for performance improvement. The open-sourced implementation of Daisen can be found at gitlab.com/akita/vis. Supplemental materials including a demo video, survey questions, evaluation study guide, and post-study evaluation survey are available at osf.io/j5ghq.

Lizhi Lin, Honglin Mu, Zenan Zhai et al.

Generative models are rapidly gaining popularity and being integrated into everyday applications, raising concerns over their safe use as various vulnerabilities are exposed. In light of this, the field of red teaming is undergoing fast-paced growth, highlighting the need for a comprehensive survey covering the entire pipeline and addressing emerging topics. Our extensive survey, which examines over 120 papers, introduces a taxonomy of fine-grained attack strategies grounded in the inherent capabilities of language models. Additionally, we have developed the "searcher" framework to unify various automatic red teaming approaches. Moreover, our survey covers novel areas including multimodal attacks and defenses, risks around LLM-based agents, overkill of harmless queries, and the balance between harmlessness and helpfulness.

Renxi Wang, Haonan Li, Xudong Han et al.

Large language models (LLMs) have achieved success in acting as agents, which interact with environments through tools such as search engines. However, LLMs are optimized for language generation instead of tool use during training or alignment, limiting their effectiveness as agents. To resolve this problem, previous work has first collected interaction trajectories between LLMs and environments, using only trajectories that successfully finished the task to fine-tune smaller models, making fine-tuning data scarce and acquiring it both difficult and costly. Discarding failed trajectories also leads to significant wastage of data and resources and limits the possible optimization paths during fine-tuning. In this paper, we argue that unsuccessful trajectories offer valuable insights, and LLMs can learn from these trajectories through appropriate quality control and fine-tuning strategies. By simply adding a prefix or suffix that tells the model whether to generate a successful trajectory during training, we improve model performance by a large margin on mathematical reasoning, multi-hop question answering, and strategic question answering tasks. We further analyze the inference results and find that our method provides a better trade-off between valuable information and errors in unsuccessful trajectories. To our knowledge, we are the first to demonstrate the value of negative trajectories and their application in agent-tunning scenarios. Our findings offer guidance for developing better agent-tuning methods and low-resource data usage techniques.

Murong Yue, Wenhan Lyu, Jennifer Suh et al.

Collaborative problem solving (CPS) is essential in mathematics education, fostering deeper learning through the exchange of ideas. Yet, classrooms often lack the resources, time, and peer dynamics needed to sustain productive CPS. Recent advancements in Large Language Models (LLMs) offer a promising avenue to enhance CPS in mathematical education. We designed and developed MathVC, a multi-persona LLM simulated virtual classroom platform to facilitate CPS in mathematics. MathVC combines a meta planning controller that monitors CPS stages-sense-making, team organization, planning, execution, validation, and predicts the next speaker, with a persona simulation stack that encodes mathematical thinking via a task schema and error-injected persona schemas seeded from teacher-specified misconceptions. We evaluated MathVC with 14 U.S. middle schoolers. Students reported constructive interaction and reaching shared solutions, describing gains in engagement, motivation, and confidence through diverse perspectives, immediate scaffolding, and human-like fallibility. Our findings also provide insights into simulating peers via LLM-based technologies for collaboration to support learning.

Yixuan Zhang, Hui Yang, Chuanchen Luo et al.

Generating realistic 3D human-object interactions (HOIs) from text descriptions is a active research topic with potential applications in virtual and augmented reality, robotics, and animation. However, creating high-quality 3D HOIs remains challenging due to the lack of large-scale interaction data and the difficulty of ensuring physical plausibility, especially in out-of-domain (OOD) scenarios. Current methods tend to focus either on the body or the hands, which limits their ability to produce cohesive and realistic interactions. In this paper, we propose OOD-HOI, a text-driven framework for generating whole-body human-object interactions that generalize well to new objects and actions. Our approach integrates a dual-branch reciprocal diffusion model to synthesize initial interaction poses, a contact-guided interaction refiner to improve physical accuracy based on predicted contact areas, and a dynamic adaptation mechanism which includes semantic adjustment and geometry deformation to improve robustness. Experimental results demonstrate that our OOD-HOI could generate more realistic and physically plausible 3D interaction pose in OOD scenarios compared to existing methods.

Zenan Ling, Longbo Li, Zhanbo Feng et al.

Deep equilibrium models (DEQs), as a typical implicit neural network, have demonstrated remarkable success on various tasks. There is, however, a lack of theoretical understanding of the connections and differences between implicit DEQs and explicit neural network models. In this paper, leveraging recent advances in random matrix theory (RMT), we perform an in-depth analysis on the eigenspectra of the conjugate kernel (CK) and neural tangent kernel (NTK) matrices for implicit DEQs, when the input data are drawn from a high-dimensional Gaussian mixture. We prove, in this setting, that the spectral behavior of these Implicit-CKs and NTKs depend on the DEQ activation function and initial weight variances, but only via a system of four nonlinear equations. As a direct consequence of this theoretical result, we demonstrate that a shallow explicit network can be carefully designed to produce the same CK or NTK as a given DEQ. Despite derived here for Gaussian mixture data, empirical results show the proposed theory and design principle also apply to popular real-world datasets.

Ma Zhuoqi, Zhang Yixuan, You Zejun et al.

Artistic image stylization aims to render the content provided by text or image with the target style, where content and style decoupling is the key to achieve satisfactory results. However, current methods for content and style disentanglement primarily rely on image supervision, which leads to two problems: 1) models can only support one modality for style or content input;2) incomplete disentanglement resulting in content leakage from the reference image. To address the above issues, this paper proposes a multimodal approach to content-style disentanglement for artistic image stylization. We construct a \textit{WikiStyle+} dataset consists of artworks with corresponding textual descriptions for style and content. Based on the multimodal dataset, we propose a disentangled representations-guided diffusion model. The disentangled representations are first learned by Q-Formers and then injected into a pre-trained diffusion model using learnable multi-step cross-attention layers. Experimental results show that our method achieves a thorough disentanglement of content and style in reference images under multimodal supervision, thereby enabling more refined stylization that aligns with the artistic characteristics of the reference style. The code of our method will be available upon acceptance.

Nurul M Suhaimi, Yixuan Zhang, Mary Joseph et al.

The world population is projected to rapidly age over the next 30 years. Given the increasing digital technology adoption amongst older adults, researchers have investigated how technology can support aging populations. However, little work has examined how technology can support older adults during crises, despite increasingly common natural disasters, public health emergencies, and other crisis scenarios in which older adults are especially vulnerable. Addressing this gap, we conducted focus groups with older adults residing in coastal locations to examine to what extent they felt technology could support them during emergencies. Our findings characterize participants' desire for tools that enhance community resilience-local knowledge, preparedness, community relationships, and communication, that help communities withstand disasters. Further, older adults' crisis technology preferences were linked to their sense of control, social relationships, and digital readiness. We discuss how a focus on community resilience can yield crisis technologies that more effectively support older adults.

Aditeya Pandey, Yixuan Zhang, John A. Guerra-Gomez et al.

In the task abstraction phase of the visualization design process, including in "design studies", a practitioner maps the observed domain goals to generalizable abstract tasks using visualization theory in order to better understand and address the users needs. We argue that this manual task abstraction process is prone to errors due to designer biases and a lack of domain background and knowledge. Under these circumstances, a collaborator can help validate and provide sanity checks to visualization practitioners during this important task abstraction stage. However, having a human collaborator is not always feasible and may be subject to the same biases and pitfalls. In this paper, we first describe the challenges associated with task abstraction. We then propose a conceptual Digital Collaborator: an artificial intelligence system that aims to help visualization practitioners by augmenting their ability to validate and reason about the output of task abstraction. We also discuss several practical design challenges of designing and implementing such systems

Yixuan Zhang, Andrea G. Parker

Food journaling is an effective method to help people identify their eating patterns and encourage healthy eating habits as it requires self-reflection on eating behaviors. Current tools have predominately focused on tracking food intake, such as carbohydrates, proteins, fats, and calories. Other factors, such as contextual information and momentary thoughts and feelings that are internal to an individual, are also essential to help people reflect upon and change attitudes about eating behaviors. However, current dietary tracking tools rarely support capturing these elements as a way to foster deep reflection. In this work, we present Eat4Thought -- a food journaling application that allows users to track their emotional, sensory, and spatio-temporal elements of meals as a means of supporting self-reflection. The application enables vivid documentation of experiences and self-reflection on the past through video recording. We describe our design process and an initial evaluation of the application. We also provide design recommendations for future work on food journaling.

Yixuan Zhang, Nurul Suhaimi, Rana Azghandi et al.

Mass emergencies increasingly pose significant threats to human life, with a disproportionate burden being incurred by older adults. Research has explored how mobile technology can mitigate the effects of mass emergencies. However, less work has examined how mobile technologies support older adults during emergencies, considering their unique needs. To address this research gap, we interviewed 16 older adults who had recent experience with an emergency evacuation to understand the perceived value of using mobile technology during emergencies. We found that there was a lack of awareness and engagement with existing crisis apps. Our findings characterize the ways in which our participants did and did not feel crisis informatics tools address human values, including basic needs and esteem needs. We contribute an understanding of how older adults used mobile technology during emergencies and their perspectives on how well such tools address human values.

Yixuan Zhang, Sara Di Bartolomeo, Fangfang Sheng et al.

Composite temporal event sequence visualizations have included sentinel event alignment techniques to cope with data volume and variety. Prior work has demonstrated the utility of using single-event alignment for understanding the precursor, co-occurring, and aftereffect events surrounding a sentinel event. However, the usefulness of single-event alignment has not been sufficiently evaluated in composite visualizations. Furthermore, recently proposed dual-event alignment techniques have not been empirically evaluated. In this work, we designed tasks around temporal event sequence and timing analysis and conducted a controlled experiment on Amazon Mechanical Turk to examine four sentinel event alignment approaches: no sentinel event alignment (NoAlign), single-event alignment (SingleAlign), dual-event alignment with left justification (DualLeft), and dual-event alignment with stretch justification (DualStretch). Differences between approaches were most pronounced with more rows of data. For understanding intermediate events between two sentinel events, dual-event alignment was the clear winner for correctness---71% vs. 18% for NoAlign and SingleAlign. For understanding the duration between two sentinel events, NoAlign was the clear winner: correctness---88% vs. 36% for DualStretch---completion time---55 seconds vs. 101 seconds for DualLeft---and error---1.5% vs. 8.4% for DualStretch. For understanding precursor and aftereffect events, there was no significant difference among approaches. A free copy of this paper, the evaluation stimuli and data, and source code are available at https://osf.io/78fs5

Zhengyuan Yang, Yixuan Zhang, Jiebo Luo

As an intuitive way of expression emotion, the animated Graphical Interchange Format (GIF) images have been widely used on social media. Most previous studies on automated GIF emotion recognition fail to effectively utilize GIF's unique properties, and this potentially limits the recognition performance. In this study, we demonstrate the importance of human related information in GIFs and conduct human-centered GIF emotion recognition with a proposed Keypoint Attended Visual Attention Network (KAVAN). The framework consists of a facial attention module and a hierarchical segment temporal module. The facial attention module exploits the strong relationship between GIF contents and human characters, and extracts frame-level visual feature with a focus on human faces. The Hierarchical Segment LSTM (HS-LSTM) module is then proposed to better learn global GIF representations. Our proposed framework outperforms the state-of-the-art on the MIT GIFGIF dataset. Furthermore, the facial attention module provides reliable facial region mask predictions, which improves the model's interpretability.

Zhongping Zhang, Yixuan Zhang, Zheng Zhou et al.

Image forgery detection is the task of detecting and localizing forged parts in tampered images. Previous works mostly focus on high resolution images using traces of resampling features, demosaicing features or sharpness of edges. However, a good detection method should also be applicable to low resolution images because compressed or resized images are common these days. To this end, we propose a Shallow Convolutional Neural Network(SCNN), capable of distinguishing the boundaries of forged regions from original edges in low resolution images. SCNN is designed to utilize the information of chroma and saturation. Based on SCNN, two approaches that are named Sliding Windows Detection (SWD) and Fast SCNN, respectively, are developed to detect and localize image forgery region. In this paper, we substantiate that Fast SCNN can detect drastic change of chroma and saturation. In image forgery detection experiments Our model is evaluated on the CASIA 2.0 dataset. The results show that Fast SCNN performs well on low resolution images and achieves significant improvements over the state-of-the-art.