Search arXiv⌕ Search

arXiv subjects

Jie Cai

Publications and source records attributed to Jie Cai.

At least 55 records · Page 3Linked to original sources

AI Trust Reshaping Administrative Burdens: Understanding Trust-Burden Dynamics in LLM-Assisted Benefits Systems

Supplemental Nutrition Assistance Program (SNAP) is an essential benefit support system provided by the US administration to 41 million federally determined low-income applicants. Through interviews with such applicants across a diverse set of experiences with the SNAP system, our findings reveal that new AI technologies like LLMs can alleviate traditional burdens but also introduce new burdens. We introduce new types of learning, compliance, and psychological costs that transform the administrative burden on applicants. We also identify how trust in AI across three dimensions--competence, integrity, and benevolence--is perceived to reduce administrative burdens, which may stem from unintended and untoward overt trust in the system. We discuss calibrating appropriate levels of user trust in LLM-based administrative systems, mitigating newly introduced burdens. In particular, our findings suggest that evidence-based information disclosure is necessary in benefits administration and propose directions for future research on trust-burden dynamics in AI-assisted administration systems.

cs.HC↗

Parental Collaboration and Closeness: Envisioning with New Couple Parents

Couples often experience a decrease in closeness as they cope with the demands of parenthood. Existing technologies have supported parenting and parental collaboration. However, these technologies do not adequately support closeness in co-parenting. We use scenarios and design probes to brainstorm with 10 new parent couples to explore and envision possibilities for technologies to support closeness. We reported parents' current technology use for co-parenting and how participants considered and envisioned co-parenting technology for closeness, including information and task sharing, emotion awareness and disclosure, and fostering fun interaction. We discuss the potential technology has for fostering closeness in co-parenting by (1) fostering interdependence by supporting parental competence and (2) integrating positive emotions and experiences, such as validation and fun, in parenting. Based on our findings, we expand the design space of technology for closeness to include interdependence. We also expand the design space for co-parenting technology by integrating more positive emotions.

cs.HC↗

Ultra-FineWeb: Efficient Data Filtering and Verification for High-Quality LLM Training Data

Data quality has become a key factor in enhancing model performance with the rapid development of large language models (LLMs). Model-driven data filtering has increasingly become a primary approach for acquiring high-quality data. However, it still faces two main challenges: (1) the lack of an efficient data verification strategy makes it difficult to provide timely feedback on data quality; and (2) the selection of seed data for training classifiers lacks clear criteria and relies heavily on human expertise, introducing a degree of subjectivity. To address the first challenge, we introduce an efficient verification strategy that enables rapid evaluation of the impact of data on LLM training with minimal computational cost. To tackle the second challenge, we build upon the assumption that high-quality seed data is beneficial for LLM training, and by integrating the proposed verification strategy, we optimize the selection of positive and negative samples and propose an efficient data filtering pipeline. This pipeline not only improves filtering efficiency, classifier quality, and robustness, but also significantly reduces experimental and inference costs. In addition, to efficiently filter high-quality data, we employ a lightweight classifier based on fastText, and successfully apply the filtering pipeline to two widely-used pre-training corpora, FineWeb and Chinese FineWeb datasets, resulting in the creation of the higher-quality Ultra-FineWeb dataset. Ultra-FineWeb contains approximately 1 trillion English tokens and 120 billion Chinese tokens. Empirical results demonstrate that the LLMs trained on Ultra-FineWeb exhibit significant performance improvements across multiple benchmark tasks, validating the effectiveness of our pipeline in enhancing both data quality and training efficiency.

cs.CL↗

A Systematic Literature Review of Infrastructure Studies in SIGCHI

Infrastructure is an indispensable part of human life. Over the past decades, the Human-Computer Interaction (HCI) community has paid increasing attention to human interactions with infrastructure. In this paper, we conducted a systematic literature review on infrastructure studies in SIGCHI, one of the most influential communities in HCI. We collected a total of 190 primary studies, covering works published between 2006 and 2024. Most of these studies are inspired by Susan Leigh Star's notion of infrastructure. We identify three major themes in infrastructure studies: growing infrastructure, appropriating infrastructure, and coping with infrastructure. Our review highlights a prevailing trend in SIGCHI's infrastructure research: a focus on informal infrastructural activities across various sociotechnical contexts. In particular, we examine studies that problematize infrastructure and alert the HCI community to its potentially harmful aspects.

cs.HC↗

VIP: Video Inpainting Pipeline for Real World Human Removal

Inpainting for real-world human and pedestrian removal in high-resolution video clips presents significant challenges, particularly in achieving high-quality outcomes, ensuring temporal consistency, and managing complex object interactions that involve humans, their belongings, and their shadows. In this paper, we introduce VIP (Video Inpainting Pipeline), a novel promptless video inpainting framework for real-world human removal applications. VIP enhances a state-of-the-art text-to-video model with a motion module and employs a Variational Autoencoder (VAE) for progressive denoising in the latent space. Additionally, we implement an efficient human-and-belongings segmentation for precise mask generation. Sufficient experimental results demonstrate that VIP achieves superior temporal consistency and visual fidelity across diverse real-world scenarios, surpassing state-of-the-art methods on challenging datasets. Our key contributions include the development of the VIP pipeline, a reference frame integration technique, and the Dual-Fusion Latent Segment Refinement method, all of which address the complexities of inpainting in long, high-resolution video sequences.

cs.CV↗

Survey on Single-Image Reflection Removal using Deep Learning Techniques

The phenomenon of reflection is quite common in digital images, posing significant challenges for various applications such as computer vision, photography, and image processing. Traditional methods for reflection removal often struggle to achieve clean results while maintaining high fidelity and robustness, particularly in real-world scenarios. Over the past few decades, numerous deep learning-based approaches for reflection removal have emerged, yielding impressive results. In this survey, we conduct a comprehensive review of the current literature by focusing on key venues such as ICCV, ECCV, CVPR, NeurIPS, etc., as these conferences and journals have been central to advances in the field. Our review follows a structured paper selection process, and we critically assess both single-stage and two-stage deep learning methods for reflection removal. The contribution of this survey is three-fold: first, we provide a comprehensive summary of the most recent work on single-image reflection removal; second, we outline task hypotheses, current deep learning techniques, publicly available datasets, and relevant evaluation metrics; and third, we identify key challenges and opportunities in deep learning-based reflection removal, highlighting the potential of this rapidly evolving research area.

cs.CV↗

Densing Law of LLMs

Large Language Models (LLMs) have emerged as a milestone in artificial intelligence, and their performance can improve as the model size increases. However, this scaling brings great challenges to training and inference efficiency, particularly for deploying LLMs in resource-constrained environments, and the scaling trend is becoming increasingly unsustainable. This paper introduces the concept of ``\textit{capacity density}'' as a new metric to evaluate the quality of the LLMs across different scales and describes the trend of LLMs in terms of both effectiveness and efficiency. To calculate the capacity density of a given target LLM, we first introduce a set of reference models and develop a scaling law to predict the downstream performance of these reference models based on their parameter sizes. We then define the \textit{effective parameter size} of the target LLM as the parameter size required by a reference model to achieve equivalent performance, and formalize the capacity density as the ratio of the effective parameter size to the actual parameter size of the target LLM. Capacity density provides a unified framework for assessing both model effectiveness and efficiency. Our further analysis of recent open-source base LLMs reveals an empirical law (the densing law)that the capacity density of LLMs grows exponentially over time. More specifically, using some widely used benchmarks for evaluation, the capacity density of LLMs doubles approximately every three months. The law provides new perspectives to guide future LLM development, emphasizing the importance of improving capacity density to achieve optimal results with minimal computational overhead.

cs.AI↗

DecorateLM: Data Engineering through Corpus Rating, Tagging, and Editing with Language Models

The performance of Large Language Models (LLMs) is substantially influenced by the pretraining corpus, which consists of vast quantities of unsupervised data processed by the models. Despite its critical role in model performance, ensuring the quality of this data is challenging due to its sheer volume and the absence of sample-level quality annotations and enhancements. In this paper, we introduce DecorateLM, a data engineering method designed to refine the pretraining corpus through data rating, tagging and editing. Specifically, DecorateLM rates texts against quality criteria, tags texts with hierarchical labels, and edits texts into a more formalized format. Due to the massive size of the pretraining corpus, adopting an LLM for decorating the entire corpus is less efficient. Therefore, to balance performance with efficiency, we curate a meticulously annotated training corpus for DecorateLM using a large language model and distill data engineering expertise into a compact 1.2 billion parameter small language model (SLM). We then apply DecorateLM to enhance 100 billion tokens of the training corpus, selecting 45 billion tokens that exemplify high quality and diversity for the further training of another 1.2 billion parameter LLM. Our results demonstrate that employing such high-quality data can significantly boost model performance, showcasing a powerful approach to enhance the quality of the pretraining corpus.

cs.CL↗

PyBench: Evaluating LLM Agent on various real-world coding tasks

The LLM Agent, equipped with a code interpreter, is capable of automatically solving real-world coding tasks, such as data analysis and image editing. However, existing benchmarks primarily focus on either simplistic tasks, such as completing a few lines of code, or on extremely complex and specific tasks at the repository level, neither of which are representative of various daily coding tasks. To address this gap, we introduce \textbf{PyBench}, a benchmark encompassing five main categories of real-world tasks, covering more than 10 types of files. Given a high-level user query and related files, the LLM Agent needs to reason and execute Python code via a code interpreter for a few turns before making a formal response to fulfill the user's requirements. Successfully addressing tasks in PyBench demands a robust understanding of various Python packages, superior reasoning capabilities, and the ability to incorporate feedback from executed code. Our evaluations indicate that current open-source LLMs are struggling with these tasks. Hence, we conduct analysis and experiments on four kinds of datasets proving that comprehensive abilities are needed for PyBench. Our fine-tuned 8B size model: \textbf{PyLlama3} achieves an exciting performance on PyBench which surpasses many 33B and 70B size models. Our Benchmark, Training Dataset, and Model are available at: {https://github.com/Mercury7353/PyBench}

cs.SE↗

MiniCPM-V: A GPT-4V Level MLLM on Your Phone

The recent surge of Multimodal Large Language Models (MLLMs) has fundamentally reshaped the landscape of AI research and industry, shedding light on a promising path toward the next AI milestone. However, significant challenges remain preventing MLLMs from being practical in real-world applications. The most notable challenge comes from the huge cost of running an MLLM with a massive number of parameters and extensive computation. As a result, most MLLMs need to be deployed on high-performing cloud servers, which greatly limits their application scopes such as mobile, offline, energy-sensitive, and privacy-protective scenarios. In this work, we present MiniCPM-V, a series of efficient MLLMs deployable on end-side devices. By integrating the latest MLLM techniques in architecture, pretraining and alignment, the latest MiniCPM-Llama3-V 2.5 has several notable features: (1) Strong performance, outperforming GPT-4V-1106, Gemini Pro and Claude 3 on OpenCompass, a comprehensive evaluation over 11 popular benchmarks, (2) strong OCR capability and 1.8M pixel high-resolution image perception at any aspect ratio, (3) trustworthy behavior with low hallucination rates, (4) multilingual support for 30+ languages, and (5) efficient deployment on mobile phones. More importantly, MiniCPM-V can be viewed as a representative example of a promising trend: The model sizes for achieving usable (e.g., GPT-4V) level performance are rapidly decreasing, along with the fast growth of end-side computation capacity. This jointly shows that GPT-4V level MLLMs deployed on end devices are becoming increasingly possible, unlocking a wider spectrum of real-world AI applications in the near future.

cs.CV↗

When Qualitative Research Meets Large Language Model: Exploring the Potential of QualiGPT as a Tool for Qualitative Coding

Qualitative research, renowned for its in-depth exploration of complex phenomena, often involves time-intensive analysis, particularly during the coding stage. Existing software for qualitative evaluation frequently lacks automatic coding capabilities, user-friendliness, and cost-effectiveness. The advent of Large Language Models (LLMs) like GPT-3 and its successors marks a transformative era for enhancing qualitative analysis. This paper introduces QualiGPT, a tool developed to address the challenges associated with using ChatGPT for qualitative analysis. Through a comparative analysis of traditional manual coding and QualiGPT's performance on both simulated and real datasets, incorporating both inductive and deductive coding approaches, we demonstrate that QualiGPT significantly improves the qualitative analysis process. Our findings show that QualiGPT enhances efficiency, transparency, and accessibility in qualitative coding. The tool's performance was evaluated using inter-rater reliability (IRR) measures, with results indicating substantial agreement between human coders and QualiGPT in various coding scenarios. In addition, we also discuss the implications of integrating AI into qualitative research workflows and outline future directions for enhancing human-AI collaboration in this field.

cs.HC↗

The Future of Learning: Large Language Models through the Lens of Students

As Large-Scale Language Models (LLMs) continue to evolve, they demonstrate significant enhancements in performance and an expansion of functionalities, impacting various domains, including education. In this study, we conducted interviews with 14 students to explore their everyday interactions with ChatGPT. Our preliminary findings reveal that students grapple with the dilemma of utilizing ChatGPT's efficiency for learning and information seeking, while simultaneously experiencing a crisis of trust and ethical concerns regarding the outcomes and broader impacts of ChatGPT. The students perceive ChatGPT as being more "human-like" compared to traditional AI. This dilemma, characterized by mixed emotions, inconsistent behaviors, and an overall positive attitude towards ChatGPT, underscores its potential for beneficial applications in education and learning. However, we argue that despite its human-like qualities, the advanced capabilities of such intelligence might lead to adverse consequences. Therefore, it's imperative to approach its application cautiously and strive to mitigate potential harms in future developments.

cs.HC↗

MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies

The burgeoning interest in developing Large Language Models (LLMs) with up to trillion parameters has been met with concerns regarding resource efficiency and practical expense, particularly given the immense cost of experimentation. This scenario underscores the importance of exploring the potential of Small Language Models (SLMs) as a resource-efficient alternative. In this context, we introduce MiniCPM, specifically the 1.2B and 2.4B non-embedding parameter variants, not only excel in their respective categories but also demonstrate capabilities on par with 7B-13B LLMs. While focusing on SLMs, our approach exhibits scalability in both model and data dimensions for future LLM research. Regarding model scaling, we employ extensive model wind tunnel experiments for stable and optimal scaling. For data scaling, we introduce a Warmup-Stable-Decay (WSD) learning rate scheduler (LRS), conducive to continuous training and domain adaptation. We present an in-depth analysis of the intriguing training dynamics that occurred in the WSD LRS. With WSD LRS, we are now able to efficiently study data-model scaling law without extensive retraining experiments on both axes of model and data, from which we derive the much higher compute optimal data-model ratio than Chinchilla Optimal. Additionally, we introduce MiniCPM family, including MiniCPM-DPO, MiniCPM-MoE and MiniCPM-128K, whose excellent performance further cementing MiniCPM's foundation in diverse SLM applications. MiniCPM models are available publicly at https://github.com/OpenBMB/MiniCPM .

cs.CL↗

Redefining Qualitative Analysis in the AI Era: Utilizing ChatGPT for Efficient Thematic Analysis

AI tools, particularly large-scale language model (LLM) based applications such as ChatGPT, have the potential to simplify qualitative research. Through semi-structured interviews with seventeen participants, we identified challenges and concerns in integrating ChatGPT into the qualitative analysis process. Collaborating with thirteen qualitative researchers, we developed a framework for designing prompts to enhance the effectiveness of ChatGPT in thematic analysis. Our findings indicate that improving transparency, providing guidance on prompts, and strengthening users' understanding of LLMs' capabilities significantly enhance the users' ability to interact with ChatGPT. We also discovered and revealed the reasons behind researchers' shift in attitude towards ChatGPT from negative to positive. This research not only highlights the importance of well-designed prompts in LLM applications but also offers reflections for qualitative researchers on the perception of AI's role. Finally, we emphasize the potential ethical risks and the impact of constructing AI ethical expectations by researchers, particularly those who are novices, on future research and AI development.

cs.HC↗

"I Upload...All Types of Different Things to Say, the World of Blindness Is More Than What They Think It Is": A Study of Blind TikTokers' Identity Work from a Flourishing Perspective

Identity work in Human-Computer Interaction (HCI) has focused on the marginalized group to explore designs to support their asset (what they have). However, little has been explored specifically on the identity work of people with disabilities, specifically, visual impairments. In this study, we interviewed 45 BlindTokers (blind users on TikTok) from various backgrounds to understand their identity work from a positive design perspective. We found that BlindTokers leverage the affordance of the platform to create positive content, share their identities, and build the community with the desire to flourish. We proposed flourishing labor to present the work conducted by BlindTokers for their community's flourishing with implications to support the flourishing labor. This work contributes to understanding blind users' experience in short video platforms and highlights that flourishing is not just an activity for any single Blind user but also a job that needs all stakeholders, including all user groups and the TikTok platform, serious and committed contribution.

cs.HC↗

Third-Party Developers and Tool Development For Community Management on Live Streaming Platform Twitch

Community management is critical for stakeholders to collaboratively build and sustain communities with socio-technical support. However, most of the existing research has mainly focused on the community members and the platform, with little attention given to the developers who act as intermediaries between the platform and community members and develop tools to support community management. This study focuses on third-party developers (TPDs) for the live streaming platform Twitch and explores their tool development practices. Using a mixed method with in-depth qualitative analysis, we found that TPDs maintain complex relationships with different stakeholders (streamers, viewers, platform, professional developers), and the multi-layered policy restricts their agency regarding idea innovation and tool development. We argue that HCI research should shift its focus from tool users to tool developers with regard to community management. We propose designs to support closer collaboration between TPDS and the platform and professional developers and streamline TPDs' development process with unified toolkits and policy documentation.

cs.HC↗

Content Moderation Justice and Fairness on Social Media: Comparisons Across Different Contexts and Platforms

Social media users may perceive moderation decisions by the platform differently, which can lead to frustration and dropout. This study investigates users' perceived justice and fairness of online moderation decisions when they are exposed to various illegal versus legal scenarios, retributive versus restorative moderation strategies, and user-moderated versus commercially moderated platforms. We conduct an online experiment on 200 American social media users of Reddit and Twitter. Results show that retributive moderation delivers higher justice and fairness for commercially moderated than for user-moderated platforms in illegal violations; restorative moderation delivers higher fairness for legal violations than illegal ones. We discuss the opportunities for platform policymaking to improve moderation system design.

cs.HC↗

A Preliminary Exploration of YouTubers' Use of Generative-AI in Content Creation

Content creators increasingly utilize generative artificial intelligence (Gen-AI) on platforms such as YouTube, TikTok, Instagram, and various blogging sites to produce imaginative images, AI-generated videos, and articles using Large Language Models (LLMs). Despite its growing popularity, there remains an underexplored area concerning the specific domains where AI-generated content is being applied, and the methodologies content creators employ with Gen-AI tools during the creation process. This study initially explores this emerging area through a qualitative analysis of 68 YouTube videos demonstrating Gen-AI usage. Our research focuses on identifying the content domains, the variety of tools used, the activities performed, and the nature of the final products generated by Gen-AI in the context of user-generated content.

cs.HC↗