arXiv subjects

Zekai Shao

Publications and source records attributed to Zekai Shao.

15 recordsLinked to original sources

arXiv2026

ChartAnno: Benchmarking Multimodal Large Language Models for Chart Annotation Generation

Annotations are essential to communicative visualization, helping explain data, emphasize key findings, and guide attention. While multimodal large language models (MLLMs) offer new opportunities for automatic chart annotation authoring, their capabilities in this task remain underexplored. To address this gap, we introduce ChartAnno, a comprehensive benchmark for evaluating MLLMs on chart annotation generation. ChartAnno contains 1,200 real-world charts with paired annotated and unannotated executable code, along with 3,600 annotation instructions spanning three levels of specificity. We also develop a multidimensional evaluation framework combining rule-based and LLM-judged metrics to assess execution, structural compliance, semantic consistency, and design effectiveness. We evaluate 10 representative MLLMs under two primary chart input settings: (1) chart code alone and (2) both code and chart image. Results reveal that proprietary models lead overall, though open-source models narrow the gap. While higher instruction specificity improves annotation quality, inferring abstract communicative intent remains difficult across all models. Providing chart images yields marginal benefit when code is available. We also examine the effect of chart code through an image-only ablation and analyze the effects of multiple task complexity indicators and instruction-level transitions. Further analyses characterize common failure modes and validate the reliability of the LLM-based judge. Experiments with D3 and SVG demonstrate the generalizability of ChartAnno beyond its primary Python setting.

cs.AI↗

arXiv2026

DVBench: Benchmarking MLLMs for Understanding Dynamic Charts and Narratives in Data Videos

While MLLMs have made significant strides in chart comprehension and video understanding, current evaluations largely isolate these capabilities, leaving a critical gap in understanding temporally evolving structured visual information. To address this gap, we introduce DVBench, a benchmark for evaluating MLLMs on data videos, a storytelling medium that integrates dynamic charts with structured narratives. We decompose data video understanding into five dimensions. DVBench comprises 300 real-world data videos and 1,000 human-verified QA pairs curated through a rigorous semi-automated pipeline. Extensive evaluations of nine MLLMs show that Gemini-3.1-Pro achieves the best overall performance, while Kimi-k2.5 is the strongest open-source model. We further identify two notable phenomena: open-source model performance does not scale strictly with parameter size, and narrative proficiency does not guarantee visual capability. Fine-grained analyses and ablation studies further reveal dimension-specific weaknesses and the effects of frame configurations and subtitle inputs, informing future MLLM development. DVBench is publicly available at https://bomiaowang.github.io/DVBench/.

cs.CL↗

arXiv2026

ChartFI: Benchmarking Faithfulness and Insightfulness of Chart Descriptions from Multimodal Large Language Models

Chart descriptions are essential for accessibility, cross-modal retrieval, and assisting readers in extracting insights from complex visualizations. As multimodal large language models (MLLMs) are increasingly adopted for automated chart description generation, a critical question arises: how faithfully and insightfully do these models actually describe charts? Current benchmarks fall short on two fronts: existing datasets consist of simple, homogeneous charts paired with shallow, fact-enumerating descriptions; and prevailing metrics fail to capture the multi-faceted nature of description quality. To address these gaps, we present the Chart Faithfulness and Insightfulness Benchmark (ChartFI-Bench). We first summarize four dimensions that characterize high-quality chart descriptions: factual accuracy, salient feature emphasis, domain-informed guidance, and chart-text complementarity. Guided by these dimensions, we construct a high-quality benchmark comprising 896 chart-description pairs, which feature visually complex charts and semantically rich descriptions. Furthermore, we design four aligned evaluation metrics -- Faithfulness, Coverage, Informativeness, and Acuity -- to systematically assess the quality of descriptions across these dimensions. Experiments conducted on mainstream MLLMs demonstrate the effectiveness of the proposed framework and reveal common weaknesses among existing models.

cs.CL↗

arXiv2026

DeepConnect: A Visual Analytics System for Bridging Interdisciplinary Research Collaborations

Interdisciplinary research collaboration is crucial for scientific innovation, but it remains difficult to initiate in practice. Existing collaborator discovery approaches are often constrained by disciplinary boundaries and static researcher profiles that do not reflect the specific context of a new collaboration goal. As a result, researchers struggle to translate open-ended collaboration goals into domain-specific tasks, evaluate candidate researchers' fit and complementarity, and establish common ground before initial contact. To address these challenges, we present DeepConnect, an LLM-augmented visual analytics system for interdisciplinary collaborator discovery. DeepConnect translates collaboration ideas into domain-specific tasks, retrieves relevant papers to ground cross-domain exploration, and provides coordinated visualizations for exploring and comparing candidate researchers. It further reveals terminology gaps and overlaps across domains and supports publication-grounded conversation rehearsal to help users prepare for outreach. We evaluate DeepConnect through two case studies, a user study, and a component-level evaluation, showing its value for complementary team formation, idea refinement, and pre-contact preparation. The DeepConnect website is available at https://deepconnect.sg.

cs.HC↗

arXiv2026

Group Orthogonal Low-Rank Adaptation for RGB-T Tracking

Parameter-efficient fine-tuning has emerged as a promising paradigm in RGB-T tracking, enabling downstream task adaptation by freezing pretrained parameters and fine-tuning only a small set of parameters. This set forms a rank space made up of multiple individual ranks, whose expressiveness directly shapes the model's adaptability. However, quantitative analysis reveals low-rank adaptation exhibits significant redundancy in the rank space, with many ranks contributing almost no practical information. This hinders the model's ability to learn more diverse knowledge to address the various challenges in RGB-T tracking. To address this issue, we propose the Group Orthogonal Low-Rank Adaptation (GOLA) framework for RGB-T tracking, which effectively leverages the rank space through structured parameter learning. Specifically, we adopt a rank decomposition partitioning strategy utilizing singular value decomposition to quantify rank importance, freeze crucial ranks to preserve the pretrained priors, and cluster the redundant ranks into groups to prepare for subsequent orthogonal constraints. We further design an inter-group orthogonal constraint strategy. This constraint enforces orthogonality between rank groups, compelling them to learn complementary features that target diverse challenges, thereby alleviating information redundancy. Experimental results demonstrate that GOLA effectively reduces parameter redundancy and enhances feature representation capabilities, significantly outperforming state-of-the-art methods across four benchmark datasets and validating its effectiveness in RGB-T tracking tasks.

cs.CV↗

arXiv2026

From $\log π$ to $π$: Taming Divergence in Soft Clipping via Bilateral Decoupled Decay of Probability Gradient Weight

Reinforcement Learning with Verifiable Rewards (RLVR) has catalyzed a leap in Large Language Model (LLM) reasoning, yet its optimization dynamics remain fragile. Standard algorithms like GRPO enforce stability via "hard clipping", which inadvertently stifles exploration by discarding gradients of tokens outside the trust region. While recent "soft clipping" methods attempt to recover these gradients, they suffer from a critical challenge: relying on log-probability gradient ($\nabla_θ\log π_θ$) yields divergent weights as probabilities vanish, destabilizing LLM training. We rethink this convention by establishing probability gradient ($\nabla_θπ_θ$) as the superior optimization primitive. Accordingly, we propose Decoupled Gradient Policy Optimization (DGPO), which employs a decoupled decay mechanism based on importance sampling ratios. By applying asymmetric, continuous decay to boundary tokens, DGPO resolves the conflict between stability and sustained exploration. Extensive experiments across DeepSeek-R1-Distill-Qwen series models (1.5B/7B/14B) demonstrate that DGPO consistently outperforms strong baselines on various mathematical benchmarks, offering a robust and scalable solution for RLVR. Our code and implementation are available at: https://github.com/FlyTune/DGPO-RL.

cs.LG↗

arXiv2026

MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning

Existing Reinforcement Learning with Verifiable Rewards (RLVR) algorithms, such as GRPO, rely on rigid, uniform, and symmetric trust region mechanisms that are fundamentally misaligned with the complex optimization dynamics of Large Language Models (LLMs). In this paper, we identify three critical challenges in these methods: (1) inefficient gradient utilization caused by the binary cutoff of hard clipping, (2) insensitive probability mass arising from uniform ratio constraints that ignore the token distribution, and (3) asymmetric signal reliability stemming from the disparate credit assignment ambiguity between positive and negative samples. To bridge these gaps, we propose Mass-Adaptive Soft Policy Optimization (MASPO), a unified framework designed to harmonize these three dimensions. MASPO integrates a differentiable soft Gaussian gating to maximize gradient utility, a mass-adaptive limiter to balance exploration across the probability spectrum, and an asymmetric risk controller to align update magnitudes with signal confidence. Extensive evaluations demonstrate that MASPO serves as a robust, all-in-one RLVR solution, significantly outperforming baselines. Our code is available at: https://github.com/FlyTune/MASPO-RL.

cs.LG↗

arXiv2026

NotebookRAG: Retrieving Multiple Notebooks to Augment the Generation of EDA Notebooks for Crowd-Wisdom

High-quality exploratory data analysis (EDA) is essential in the data science pipeline, but remains highly dependent on analysts' expertise and effort. While recent LLM-based approaches partially reduce this burden, they struggle to generate effective analysis plans and appropriate insights and visualizations when user intent is abstract. Meanwhile, a vast collection of analysis notebooks produced across platforms and organizations contains rich analytical knowledge that can potentially guide automated EDA. Retrieval-augmented generation (RAG) provides a natural way to leverage such corpora, but general methods often treat notebooks as static documents and fail to fully exploit their potential knowledge for automating EDA. To address these limitations, we propose NotebookRAG, a method that takes user intent, datasets, and existing notebooks as input to retrieve, enhance, and reuse relevant notebook content for automated EDA generation. For retrieval, we transform code cells into context-enriched executable components, which improve retrieval quality and enable rerun with new data to generate updated visualizations and reliable insights. For generation, an agent leverages enhanced retrieval content to construct effective EDA plans, derive insights, and produce appropriate visualizations. Evidence from a user study with 24 participants confirms the superiority of our method in producing high-quality and intent-aligned EDA notebooks.

cs.HC↗

arXiv2025

Do Language Model Agents Align with Humans in Rating Visualizations? An Empirical Study

Large language models encode knowledge in various domains and demonstrate the ability to understand visualizations. They may also capture visualization design knowledge and potentially help reduce the cost of formative studies. However, it remains a question whether large language models are capable of predicting human feedback on visualizations. To investigate this question, we conducted three studies to examine whether large model-based agents can simulate human ratings in visualization tasks. The first study, replicating a published study involving human subjects, shows agents are promising in conducting human-like reasoning and rating, and its result guides the subsequent experimental design. The second study repeated six human-subject studies reported in literature on subjective ratings, but replacing human participants with agents. Consulting with five human experts, this study demonstrates that the alignment of agent ratings with human ratings positively correlates with the confidence levels of the experts before the experiments. The third study tests commonly used techniques for enhancing agents, including preprocessing visual and textual inputs, and knowledge injection. The results reveal the issues of these techniques in robustness and potential induction of biases. The three studies indicate that language model-based agents can potentially simulate human ratings in visualization experiments, provided that they are guided by high-confidence hypotheses from expert evaluators. Additionally, we demonstrate the usage scenario of swiftly evaluating prototypes with agents. We discuss insights and future directions for evaluating and improving the alignment of agent ratings with human ratings. We note that simulation may only serve as complements and cannot replace user studies.

cs.HC↗

arXiv2025

Unlocking Scientific Concepts: How Effective Are LLM-Generated Analogies for Student Understanding and Classroom Practice?

Teaching scientific concepts is essential but challenging, and analogies help students connect new concepts to familiar ideas. Advancements in large language models (LLMs) enable generating analogies, yet their effectiveness in education remains underexplored. In this paper, we first conducted a two-stage study involving high school students and teachers to assess the effectiveness of LLM-generated analogies in biology and physics through a controlled in-class test and a classroom field study. Test results suggested that LLM-generated analogies could enhance student understanding particularly in biology, but require teachers' guidance to prevent over-reliance and overconfidence. Classroom experiments suggested that teachers could refine LLM-generated analogies to their satisfaction and inspire new analogies from generated ones, encouraged by positive classroom feedback and homework performance boosts. Based on findings, we developed and evaluated a practical system to help teachers generate and refine teaching analogies. We discussed future directions for developing and evaluating LLM-supported teaching and learning by analogy.

cs.HC↗

arXiv2025

ChartInsighter: An Approach for Mitigating Hallucination in Time-series Chart Summary Generation with A Benchmark Dataset

Effective chart summary can significantly reduce the time and effort decision makers spend interpreting charts, enabling precise and efficient communication of data insights. Previous studies have faced challenges in generating accurate and semantically rich summaries of time-series data charts. In this paper, we identify summary elements and common hallucination types in the generation of time-series chart summaries, which serve as our guidelines for automatic generation. We introduce ChartInsighter, which automatically generates chart summaries of time-series data, effectively reducing hallucinations in chart summary generation. Specifically, we assign multiple agents to generate the initial chart summary and collaborate iteratively, during which they invoke external data analysis modules to extract insights and compile them into a coherent summary. Additionally, we implement a self-consistency test method to validate and correct our summary. We create a high-quality benchmark of charts and summaries, with hallucination types annotated on a sentence-by-sentence basis, facilitating the evaluation of the effectiveness of reducing hallucinations. Our evaluations using our benchmark show that our method surpasses state-of-the-art models, and that our summary hallucination rate is the lowest, which effectively reduces various hallucinations and improves summary quality. The benchmark is available at https://github.com/wangfen01/ChartInsighter.

cs.CL↗

arXiv2025

Narrative Player: Reviving Data Narratives with Visuals

Data-rich documents are commonly found across various fields such as business, finance, and science. However, a general limitation of these documents for reading is their reliance on text to convey data and facts. Visual representation of text aids in providing a satisfactory reading experience in comprehension and engagement. However, existing work emphasizes presenting the insights of local text context, rather than fully conveying data stories within the whole paragraphs and engaging readers. To provide readers with satisfactory data stories, this paper presents Narrative Player, a novel method that automatically revives data narratives with consistent and contextualized visuals. Specifically, it accepts a paragraph and corresponding data table as input and leverages LLMs to characterize the clauses and extract contextualized data facts. Subsequently, the facts are transformed into a coherent visualization sequence with a carefully designed optimization-based approach. Animations are also assigned between adjacent visualizations to enable seamless transitions. Finally, the visualization sequence, transition animations, and audio narration generated by text-to-speech technologies are rendered into a data video. The evaluation results showed that the automatic-generated data videos were well-received by participants and experts for enhancing reading.

cs.HC↗

arXiv2024

Fine-Tuned Large Language Model for Visualization System: A Study on Self-Regulated Learning in Education

Large Language Models (LLMs) have shown great potential in intelligent visualization systems, especially for domain-specific applications. Integrating LLMs into visualization systems presents challenges, and we categorize these challenges into three alignments: domain problems with LLMs, visualization with LLMs, and interaction with LLMs. To achieve these alignments, we propose a framework and outline a workflow to guide the application of fine-tuned LLMs to enhance visual interactions for domain-specific tasks. These alignment challenges are critical in education because of the need for an intelligent visualization system to support beginners' self-regulated learning. Therefore, we apply the framework to education and introduce Tailor-Mind, an interactive visualization system designed to facilitate self-regulated learning for artificial intelligence beginners. Drawing on insights from a preliminary study, we identify self-regulated learning tasks and fine-tuning objectives to guide visualization design and tuning data construction. Our focus on aligning visualization with fine-tuned LLM makes Tailor-Mind more like a personalized tutor. Tailor-Mind also supports interactive recommendations to help beginners better achieve their learning goals. Model performance evaluations and user studies confirm that Tailor-Mind improves the self-regulated learning experience, effectively validating the proposed framework.

cs.HC↗

arXiv2024

LEVA: Using Large Language Models to Enhance Visual Analytics

Visual analytics supports data analysis tasks within complex domain problems. However, due to the richness of data types, visual designs, and interaction designs, users need to recall and process a significant amount of information when they visually analyze data. These challenges emphasize the need for more intelligent visual analytics methods. Large language models have demonstrated the ability to interpret various forms of textual data, offering the potential to facilitate intelligent support for visual analytics. We propose LEVA, a framework that uses large language models to enhance users' VA workflows at multiple stages: onboarding, exploration, and summarization. To support onboarding, we use large language models to interpret visualization designs and view relationships based on system specifications. For exploration, we use large language models to recommend insights based on the analysis of system status and data to facilitate mixed-initiative exploration. For summarization, we present a selective reporting strategy to retrace analysis history through a stream visualization and generate insight reports with the help of large language models. We demonstrate how LEVA can be integrated into existing visual analytics systems. Two usage scenarios and a user study suggest that LEVA effectively aids users in conducting visual analytics.

cs.HC↗

arXiv2023

A Comprehensive Capability Analysis of GPT-3 and GPT-3.5 Series Models

GPT series models, such as GPT-3, CodeX, InstructGPT, ChatGPT, and so on, have gained considerable attention due to their exceptional natural language processing capabilities. However, despite the abundance of research on the difference in capabilities between GPT series models and fine-tuned models, there has been limited attention given to the evolution of GPT series models' capabilities over time. To conduct a comprehensive analysis of the capabilities of GPT series models, we select six representative models, comprising two GPT-3 series models (i.e., davinci and text-davinci-001) and four GPT-3.5 series models (i.e., code-davinci-002, text-davinci-002, text-davinci-003, and gpt-3.5-turbo). We evaluate their performance on nine natural language understanding (NLU) tasks using 21 datasets. In particular, we compare the performance and robustness of different models for each task under zero-shot and few-shot scenarios. Our extensive experiments reveal that the overall ability of GPT series models on NLU tasks does not increase gradually as the models evolve, especially with the introduction of the RLHF training strategy. While this strategy enhances the models' ability to generate human-like responses, it also compromises their ability to solve some tasks. Furthermore, our findings indicate that there is still room for improvement in areas such as model robustness.

cs.CL↗