arXiv subjects

Xinyu Li

Publications and source records attributed to Xinyu Li.

At least 127 records · Page 7Linked to original sources

arXiv2025

Unraveling and Mitigating Retriever Inconsistencies in Retrieval-Augmented Large Language Models

Although Retrieval-Augmented Large Language Models (RALMs) demonstrate their superiority in terms of factuality, they do not consistently outperform the original retrieval-free Language Models (LMs). Our experiments reveal that this example-level performance inconsistency exists not only between retrieval-augmented and retrieval-free LM but also among different retrievers. To understand this phenomenon, we investigate the degeneration behavior of RALMs and theoretically decompose it into four categories. Further analysis based on our decomposition reveals that the innate difference in knowledge sources and the unpredictable degeneration of the reader model contribute most to the inconsistency. Drawing from our analysis, we introduce Ensemble of Retrievers (EoR), a trainable framework that can adaptively retrieve from different knowledge sources and effectively decrease unpredictable reader errors. Our experiments on Open Domain Question Answering show that EoR substantially improves performance over the RALM with a single retriever by considerably reducing inconsistent behaviors.

cs.AI↗

arXiv2025

Interpretable Concept-based Deep Learning Framework for Multimodal Human Behavior Modeling

In the contemporary era of intelligent connectivity, Affective Computing (AC), which enables systems to recognize, interpret, and respond to human behavior states, has become an integrated part of many AI systems. As one of the most critical components of responsible AI and trustworthiness in all human-centered systems, explainability has been a major concern in AC. Particularly, the recently released EU General Data Protection Regulation requires any high-risk AI systems to be sufficiently interpretable, including biometric-based systems and emotion recognition systems widely used in the affective computing field. Existing explainable methods often compromise between interpretability and performance. Most of them focus only on highlighting key network parameters without offering meaningful, domain-specific explanations to the stakeholders. Additionally, they also face challenges in effectively co-learning and explaining insights from multimodal data sources. To address these limitations, we propose a novel and generalizable framework, namely the Attention-Guided Concept Model (AGCM), which provides learnable conceptual explanations by identifying what concepts that lead to the predictions and where they are observed. AGCM is extendable to any spatial and temporal signals through multimodal concept alignment and co-learning, empowering stakeholders with deeper insights into the model's decision-making process. We validate the efficiency of AGCM on well-established Facial Expression Recognition benchmark datasets while also demonstrating its generalizability on more complex real-world human behavior understanding applications.

cs.CV↗

arXiv2025

Intrinsic Gaussian Process Regression Modeling for Manifold-valued Response Variable

Extrinsic Gaussian process regression methods, such as wrapped Gaussian process, have been developed to analyze manifold data. However, there is a lack of intrinsic Gaussian process methods for studying complex data with manifold-valued response variables. In this paper, we first apply the parallel transport operator on Riemannian manifold to propose an intrinsic covariance structure that addresses a critical aspect of constructing a well-defined Gaussian process regression model. We then propose a novel intrinsic Gaussian process regression model for manifold-valued data, which can be applied to data situated not only on Euclidean submanifolds but also on manifolds without a natural ambient space. We establish the asymptotic properties of the proposed models, including information consistency and posterior consistency, and we also show that the posterior distribution of the regression function is invariant to the choice of orthonormal frames for the coordinate representations of the covariance function. Numerical studies, including simulation and real examples, indicate that the proposed methods work well.

stat.ML↗

arXiv2025

TeamVision: An AI-powered Learning Analytics System for Supporting Reflection in Team-based Healthcare Simulation

Healthcare simulations help learners develop teamwork and clinical skills in a risk-free setting, promoting reflection on real-world practices through structured debriefs. However, despite video's potential, it is hard to use, leaving a gap in providing concise, data-driven summaries for supporting effective debriefing. Addressing this, we present TeamVision, an AI-powered multimodal learning analytics (MMLA) system that captures voice presence, automated transcriptions, body rotation, and positioning data, offering educators a dashboard to guide debriefs immediately after simulations. We conducted an in-the-wild study with 56 teams (221 students) and recorded debriefs led by six teachers using TeamVision. Follow-up interviews with 15 students and five teachers explored perceptions of its usefulness, accuracy, and trustworthiness. This paper examines: i) how TeamVision was used in debriefing, ii) what educators found valuable and challenging, and iii) perceptions of its effectiveness. Results suggest TeamVision enables flexible debriefing and highlights the challenges and implications of using AI-powered systems in healthcare simulation.

cs.HC↗

arXiv2024

On the minimal parabolic induction

Motivated by Beilinson-Bernstein's proof of the Jantzen conjectures, we define the minimal parabolic induction functor for Kac-Moody algebras, and establish some basic properties. As applications of the formal theory, we examine first extension groups between simple highest weight modules in the category of weight modules, and analyze the annihilators of some simple highest weight modules.

math.RT↗

arXiv2024

NowYouSee Me: Context-Aware Automatic Audio Description

Audio Description (AD) plays a pivotal role as an application system aimed at guaranteeing accessibility in multimedia content, which provides additional narrations at suitable intervals to describe visual elements, catering specifically to the needs of visually impaired audiences. In this paper, we introduce $\mathrm{CA^3D}$, the pioneering unified Context-Aware Automatic Audio Description system that provides AD event scripts with precise locations in the long cinematic content. Specifically, $\mathrm{CA^3D}$ system consists of: 1) a Temporal Feature Enhancement Module to efficiently capture longer term dependencies, 2) an anchor-based AD event detector with feature suppression module that localizes the AD events and extracts discriminative feature for AD generation, and 3) a self-refinement module that leverages the generated output to tweak AD event boundaries from coarse to fine. Unlike conventional methods which rely on metadata and ground truth AD timestamp for AD detection and generation tasks, the proposed $\mathrm{CA^3D}$ is the first end-to-end trainable system that only uses visual cue. Extensive experiments demonstrate that the proposed $\mathrm{CA^3D}$ improves existing architectures for both AD event detection and script generation metrics, establishing the new state-of-the-art performances in the AD automation.

cs.CV↗

arXiv2024

GANPrompt: Enhancing Robustness in LLM-Based Recommendations with GAN-Enhanced Diversity Prompts

In recent years, Large Language Models (LLMs) have demonstrated remarkable proficiency in comprehending and generating natural language, with a growing prevalence in the domain of recommendation systems. However, LLMs still face a significant challenge called prompt sensitivity, which refers to that it is highly susceptible to the influence of prompt words. This inconsistency in response to minor alterations in prompt input may compromise the accuracy and resilience of recommendation models. To address this issue, this paper proposes GANPrompt, a multi-dimensional LLMs prompt diversity framework based on Generative Adversarial Networks (GANs). The framework enhances the model's adaptability and stability to diverse prompts by integrating GANs generation techniques with the deep semantic understanding capabilities of LLMs. GANPrompt first trains a generator capable of producing diverse prompts by analysing multidimensional user behavioural data. These diverse prompts are then used to train the LLMs to improve its performance in the face of unseen prompts. Furthermore, to ensure a high degree of diversity and relevance of the prompts, this study introduces a mathematical theory-based diversity constraint mechanism that optimises the generated prompts to ensure that they are not only superficially distinct, but also semantically cover a wide range of user intentions. Through extensive experiments on multiple datasets, we demonstrate the effectiveness of the proposed framework, especially in improving the adaptability and robustness of recommendation systems in complex and dynamic environments. The experimental results demonstrate that GANPrompt yields substantial enhancements in accuracy and robustness relative to existing state-of-the-art methodologies.

cs.IR↗

arXiv2024

Beware of Metacognitive Laziness: Effects of Generative Artificial Intelligence on Learning Motivation, Processes, and Performance

With the continuous development of technological and educational innovation, learners nowadays can obtain a variety of support from agents such as teachers, peers, education technologies, and recently, generative artificial intelligence such as ChatGPT. The concept of hybrid intelligence is still at a nascent stage, and how learners can benefit from a symbiotic relationship with various agents such as AI, human experts and intelligent learning systems is still unknown. The emerging concept of hybrid intelligence also lacks deep insights and understanding of the mechanisms and consequences of hybrid human-AI learning based on strong empirical research. In order to address this gap, we conducted a randomised experimental study and compared learners' motivations, self-regulated learning processes and learning performances on a writing task among different groups who had support from different agents (ChatGPT, human expert, writing analytics tools, and no extra tool). A total of 117 university students were recruited, and their multi-channel learning, performance and motivation data were collected and analysed. The results revealed that: learners who received different learning support showed no difference in post-task intrinsic motivation; there were significant differences in the frequency and sequences of the self-regulated learning processes among groups; ChatGPT group outperformed in the essay score improvement but their knowledge gain and transfer were not significantly different. Our research found that in the absence of differences in motivation, learners with different supports still exhibited different self-regulated learning processes, ultimately leading to differentiated performance. What is particularly noteworthy is that AI technologies such as ChatGPT may promote learners' dependence on technology and potentially trigger metacognitive laziness.

cs.AI↗

arXiv2024

Self-regulated Learning Processes in Secondary Education: A Network Analysis of Trace-based Measures

While the capacity to self-regulate has been found to be crucial for secondary school students, prior studies often rely on self-report surveys and think-aloud protocols that present notable limitations in capturing self-regulated learning (SRL) processes. This study advances the understanding of SRL in secondary education by using trace data to examine SRL processes during multi-source writing tasks, with higher education participants included for comparison. We collected fine-grained trace data from 66 secondary school students and 59 university students working on the same writing tasks within a shared SRL-oriented learning environment. The data were labelled using Bannert's validated SRL coding scheme to reflect specific SRL processes, and we examined the relationship between these processes, essay performance, and educational levels. Using epistemic network analysis (ENA) to model and visualise the interconnected SRL processes in Bannert's coding scheme, we found that: (a) secondary school students predominantly engaged in three SRL processes -- Orientation, Re-reading, and Elaboration/Organisation; (b) high-performing secondary students engaged more in Re-reading, while low-performing students showed more Orientation process; and (c) higher education students exhibited more diverse SRL processes such as Monitoring and Evaluation than their secondary education counterparts, who heavily relied on following task instructions and rubrics to guide their writing. These findings highlight the necessity of designing scaffolding tools and developing teacher training programs to enhance awareness and development of SRL skills for secondary school learners.

cs.HC↗

arXiv2024

The FLoRA Engine: Using Analytics to Measure and Facilitate Learners' own Regulation Activities

The focus of education is increasingly set on learners' ability to regulate their own learning within technology-enhanced learning environments (TELs). Prior research has shown that self-regulated learning (SRL) leads to better learning performance. However, many learners struggle to self-regulate their learning productively, as they typically need to navigate a myriad of cognitive, metacognitive, and motivational processes that SRL demands. To address these challenges, the FLoRA engine is developed to assist students, workers, and professionals in improving their SRL skills and becoming productive lifelong learners. FLoRA incorporates several learning tools that are grounded in SRL theory and enhanced with learning analytics (LA), aimed at improving learners' mastery of different SRL skills. The engine tracks learners' SRL behaviours during a learning task and provides automated scaffolding to help learners effectively regulate their learning. The main contributions of FLoRA include (1) creating instrumentation tools that unobtrusively collect intensively sampled, fine-grained, and temporally ordered trace data about learners' learning actions, (2) building a trace parser that uses LA and related analytical technique (e.g., process mining) to model and understand learners' SRL processes, and (3) providing a scaffolding module that presents analytics-based adaptive, personalised scaffolds based on students' learning progress. The architecture and implementation of the FLoRA engine are also discussed in this paper.

cs.HC↗

arXiv2024

GEXIA: Granularity Expansion and Iterative Approximation for Scalable Multi-grained Video-language Learning

In various video-language learning tasks, the challenge of achieving cross-modality alignment with multi-grained data persists. We propose a method to tackle this challenge from two crucial perspectives: data and modeling. Given the absence of a multi-grained video-text pretraining dataset, we introduce a Granularity EXpansion (GEX) method with Integration and Compression operations to expand the granularity of a single-grained dataset. To better model multi-grained data, we introduce an Iterative Approximation Module (IAM), which embeds multi-grained videos and texts into a unified, low-dimensional semantic space while preserving essential information for cross-modal alignment. Furthermore, GEXIA is highly scalable with no restrictions on the number of video-text granularities for alignment. We evaluate our work on three categories of video tasks across seven benchmark datasets, showcasing state-of-the-art or comparable performance. Remarkably, our model excels in tasks involving long-form video understanding, even though the pretraining dataset only contains short video clips.

cs.CV↗

arXiv2024

Personalized Language Modeling from Personalized Human Feedback

Personalized large language models (LLMs) are designed to tailor responses to individual user preferences. While Reinforcement Learning from Human Feedback (RLHF) is a commonly used framework for aligning LLMs with human preferences, vanilla RLHF assumes that all human preferences share the same distribution, preventing fine-tuned LLMs from generating personalized content when user preferences are diverse. In this work, we propose Personalized-RLHF (P-RLHF), an efficient framework that utilizes a lightweight user model to capture individual user preferences and jointly learns the user model and the personalized LLM from human feedback. P-RLHF exhibits the following three characteristics: (1) It enables an LLM to generate personalized content and scale efficiently with growing number of users. (2) It handles both explicit user preferences described as textual input and implicit user preferences encoded in the feedback data. (3) It eliminates the need for users to fully articulate their preferences, which are normally needed for prompting LLMs to generate personalized content yet are often impractical to obtain in real-world scenarios. Our experimental results show that personalized LLMs trained using P-RLHF generate responses that are more closely aligned with individual user preferences, outperforming vanilla, non-personalized RLHF and prompting-based personalization approaches across different tasks. We opensource our code at https://github.com/HumainLab/Personalized_RLHF.

cs.CL↗

arXiv2024

Transportation Marketplace Rate Forecast Using Signature Transform

Freight transportation marketplace rates are typically challenging to forecast accurately. In this work, we have developed a novel statistical technique based on signature transforms and have built a predictive and adaptive model to forecast these marketplace rates. Our technique is based on two key elements of the signature transform: one being its universal nonlinearity property, which linearizes the feature space and hence translates the forecasting problem into linear regression, and the other being the signature kernel, which allows for comparing computationally efficiently similarities between time series data. Combined, it allows for efficient feature generation and precise identification of seasonality and regime switching in the forecasting process. An algorithm based on our technique has been deployed by Amazon trucking operations, with far superior forecast accuracy and better interpretability versus commercially available industry models, even during the COVID-19 pandemic and the Ukraine conflict. Furthermore, our technique is able to capture the influence of business cycles and the heterogeneity of the marketplace, improving prediction accuracy by more than fivefold, with an estimated annualized saving of \$50MM.

cs.LG↗

arXiv2024

Blockchain-Based Ad Auctions and Bayesian Persuasion: An Analysis of Advertiser Behavior

This paper explores how ad platforms can utilize Bayesian persuasion within blockchain-based auction systems to strategically influence advertiser behavior despite increased transparency. By integrating game-theoretic models with machine learning techniques and the principles of blockchain technology, we analyze the role of strategic information disclosure in ad auctions. Our findings demonstrate that even in environments with inherent transparency, ad platforms can design signals to affect advertisers' beliefs and bidding strategies. A detailed case study illustrates how machine learning can predict advertiser responses to different signals, leading to optimized signaling strategies that increase expected revenue. The study contributes to the literature by extending Bayesian persuasion models to transparent systems and providing practical insights for auction design in the digital advertising industry.

econ.GN↗

arXiv2024

Chatting with a Learning Analytics Dashboard: The Role of Generative AI Literacy on Learner Interaction with Conventional and Scaffolding Chatbots

Learning analytics dashboards (LADs) simplify complex learner data into accessible visualisations, providing actionable insights for educators and students. However, their educational effectiveness has not always matched the sophistication of the technology behind them. Explanatory and interactive LADs, enhanced by generative AI (GenAI) chatbots, hold promise by enabling dynamic, dialogue-based interactions with data visualisations and offering personalised feedback through text. Yet, the effectiveness of these tools may be limited by learners' varying levels of GenAI literacy, a factor that remains underexplored in current research. This study investigates the role of GenAI literacy in learner interactions with conventional (reactive) versus scaffolding (proactive) chatbot-assisted LADs. Through a comparative analysis of 81 participants, we examine how GenAI literacy is associated with learners' ability to interpret complex visualisations and their cognitive processes during interactions with chatbot-assisted LADs. Results show that while both chatbots significantly improved learner comprehension, those with higher GenAI literacy benefited the most, particularly with conventional chatbots, demonstrating diverse prompting strategies. Findings highlight the importance of considering learners' GenAI literacy when integrating GenAI chatbots in LADs and educational technologies. Incorporating scaffolding techniques within GenAI chatbots can be an effective strategy, offering a more guided experience that reduces reliance on learners' GenAI literacy.

cs.HC↗

arXiv2024

The Effects of Generative AI Agents and Scaffolding on Enhancing Students' Comprehension of Visual Learning Analytics

Visual learning analytics (VLA) is becoming increasingly adopted in educational technologies and learning analytics dashboards to convey critical insights to students and educators. Yet many students experienced difficulties in comprehending complex VLA due to their limited data visualisation literacy. While conventional scaffolding approaches like data storytelling have shown effectiveness in enhancing students' comprehension of VLA, these approaches remain difficult to scale and adapt to individual learning needs. Generative AI (GenAI) technologies, especially conversational agents, offer potential solutions by providing personalised and dynamic support to enhance students' comprehension of VLA. This study investigates the effectiveness of GenAI agents, particularly when integrated with scaffolding techniques, in improving students' comprehension of VLA. A randomised controlled trial was conducted with 117 higher education students to compare the effects of two types of GenAI agents: passive agents, which respond to student queries, and proactive agents, which utilise scaffolding questions, against standalone scaffolding in a VLA comprehension task. The results show that passive agents yield comparable improvements to standalone scaffolding both during and after the intervention. Notably, proactive GenAI agents significantly enhance students' VLA comprehension compared to both passive agents and standalone scaffolding, with these benefits persisting beyond the intervention. These findings suggest that integrating GenAI agents with scaffolding can have lasting positive effects on students' comprehension skills and support genuine learning.

cs.HC↗

arXiv2024

Video Token Merging for Long-form Video Understanding

As the scale of data and models for video understanding rapidly expand, handling long-form video input in transformer-based models presents a practical challenge. Rather than resorting to input sampling or token dropping, which may result in information loss, token merging shows promising results when used in collaboration with transformers. However, the application of token merging for long-form video processing is not trivial. We begin with the premise that token merging should not rely solely on the similarity of video tokens; the saliency of tokens should also be considered. To address this, we explore various video token merging strategies for long-form video classification, starting with a simple extension of image token merging, moving to region-concentrated merging, and finally proposing a learnable video token merging (VTM) algorithm that dynamically merges tokens based on their saliency. Extensive experimental results show that we achieve better or comparable performances on the LVU, COIN, and Breakfast datasets. Moreover, our approach significantly reduces memory costs by 84% and boosts throughput by approximately 6.89 times compared to baseline algorithms.

cs.CV↗

arXiv2024

Leveraging Connected Vehicle Data for Near-Crash Detection and Analysis in Urban Environments

Urban traffic safety is a pressing concern in modern transportation systems, especially in rapidly growing metropolitan areas where increased traffic congestion, complex road networks, and diverse driving behaviors exacerbate the risk of traffic incidents. Traditional traffic crash data analysis offers valuable insights but often overlooks a broader range of road safety risks. Near-crash events, which occur more frequently and signal potential collisions, provide a more comprehensive perspective on traffic safety. However, city-scale analysis of near-crash events remains limited due to the significant challenges in large-scale real-world data collection, processing, and analysis. This study utilizes one month of connected vehicle data, comprising billions of records, to detect and analyze near-crash events across the road network in the City of San Antonio, Texas. We propose an efficient framework integrating spatial-temporal buffering and heading algorithms to accurately identify and map near-crash events. A binary logistic regression model is employed to assess the influence of road geometry, traffic volume, and vehicle types on near-crash risks. Additionally, we examine spatial and temporal patterns, including variations by time of day, day of the week, and road category. The findings of this study show that the vehicles on more than half of road segments will be involved in at least one near-crash event. In addition, more than 50% near-crash events involved vehicles traveling at speeds over 57.98 mph, and many occurred at short distances between vehicles. The analysis also found that wider roadbeds and multiple lanes reduced near-crash risks, while single-unit trucks slightly increased the likelihood of near-crash events. Finally, the spatial-temporal analysis revealed that near-crash risks were most prominent during weekday peak hours, especially in downtown areas.

stat.AP↗