Dazhen Deng

KEditVis: A Visual Analytics System for Knowledge Editing of Large Language Models

Zhenning Chen, Hanbei Zhan, Yanwei Huang, Xin Wu et autres

Large Language Models (LLMs) demonstrate exceptional capabilities in factual question answering, yet they sometimes provide incorrect responses. To address this issue, knowledge editing techniques have emerged as effective methods for correcting factual information in LLMs. However, typical knowledge editing workflows struggle with …

cn, hk (code pays fourni par la source)

A Declarative Grammar for Interactive Trajectory Visualization: Interaction as First-Class Component

Shifu Chen, Xiaodan Miao, Dazhen Deng, Zikun Deng et autres

cn (code pays fourni par la source)

GraphTide: Augmenting Knowledge-Intensive Text with Progressive Nested Graph

Xin Qian, Dazhen Deng, Zhaoping He, Xingbo Wang et autres

Knowledge-intensive text usually contains fruitful entities and complex relationships, such as academic articles and scientific exposition. Reading and comprehending such texts often demands considerable time and mental effort to track the relationships between entities. To reduce the burden, we present GraphTide, a …

GraphTide: Augmenting Knowledge-Intensive Text with Progressive Nested Graph

Xin Qian, Dazhen Deng, Zhaoping He, Xingbo Wang et autres

Knowledge-intensive text usually contains fruitful entities and complex relationships, such as academic articles and scientific exposition. Reading and comprehending such texts often demands considerable time and mental effort to track the relationships between entities. To reduce the burden, we present GraphTide, a …

Making Multimodal LLMs Reliable Chart Data Extractors: A Benchmark and Training Framework

Peizhi Ying, Liqi Cheng, Kuilin Peng, Dazhen Deng et autres

Chart data extraction, which reverse-engineers data tables from chart images, is essential for reproducibility, analysis, retrieval, and redesign. Existing interactive tools are reliable but tedious, and mixed-initiative systems, while more efficient, lack generalizability. Recent multimodal large language models (MLLMs) offer a unified …

cn (code pays fourni par la source)

NoteFlow: Leveraging Charts as Sight Glasses for Consistent and Continuous Data Flow Tracing

Dazhen Deng, Huawei Zheng, Kai Xiong, Xinjing Yi et autres

Computational notebooks offer a flexible environment for exploratory data analysis (EDA), but this flexibility often leads to disorganized and iterative execution of notebook cells, making it difficult to track how data states evolve. Consequently, data scientists must devote extra mental effort to …

cn (code pays fourni par la source)

KEditVis: A Visual Analytics System for Knowledge Editing of Large Language Models

Zhenning Chen, Hanbei Zhan, Yanwei Huang, Xin Wu et autres

Large Language Models (LLMs) demonstrate exceptional capabilities in factual question answering, yet they sometimes provide incorrect responses. To address this issue, knowledge editing techniques have emerged as effective methods for correcting factual information in LLMs. However, typical knowledge editing workflows struggle with …

KEditVis: A Visual Analytics System for Knowledge Editing of Large Language Models

Zhenning Chen, Hanbei Zhan, Yanwei Huang, Xin Wu et autres

Large Language Models (LLMs) demonstrate exceptional capabilities in factual question answering, yet they sometimes provide incorrect responses. To address this issue, knowledge editing techniques have emerged as effective methods for correcting factual information in LLMs. However, typical knowledge editing workflows struggle with …

cn (code pays fourni par la source)

StealthGraph: Exposing Domain-Specific Risks in LLMs through Knowledge-Graph-Guided Harmful Prompt Generation

Huawei Zheng, Xinqi Jiang, Sen Yang, Shouling Ji et autres

Large language models (LLMs) are increasingly applied in specialized domains such as finance and healthcare, where they introduce unique safety risks. Domain-specific datasets of harmful prompts remain scarce and still largely rely on manual construction; public datasets mainly focus on explicit harmful …

StealthGraph: Exposing Domain-Specific Risks in LLMs through Knowledge-Graph-Guided Harmful Prompt Generation

Huawei Zheng, Xinqi Jiang, Sen Yang, Shouling Ji et autres

Large language models (LLMs) are increasingly applied in specialized domains such as finance and healthcare, where they introduce unique safety risks. Domain-specific datasets of harmful prompts remain scarce and still largely rely on manual construction; public datasets mainly focus on explicit harmful …

cn (code pays fourni par la source)

CycleChart: A Unified Consistency-Based Learning Framework for Bidirectional Chart Understanding and Generation

Dazhen Deng, Sen Yang, Yuchen He, Yuan Tian et autres

Current chart-related tasks, such as chart generation (NL2Chart), chart schema parsing, chart data parsing, and chart question answering (ChartQA), are typically studied in isolation, preventing models from learning the shared semantics that link chart creation and interpretation. We introduce CycleChart, a consistency-based …

CycleChart: A Unified Consistency-Based Learning Framework for Bidirectional Chart Understanding and Generation

Dazhen Deng, Sen Yang, Yuchen He, Yuan Tian et autres

Current chart-related tasks, such as chart generation (NL2Chart), chart schema parsing, chart data parsing, and chart question answering (ChartQA), are typically studied in isolation, preventing models from learning the shared semantics that link chart creation and interpretation. We introduce CycleChart, a consistency-based …

cn (code pays fourni par la source)