20 papers
Why Generate a Click? Plug-in Decision Heads for Web Agents
Chengguang Gan, Yunhao Liang, Hanjun Wei +3
Web agents built on vision language models decide every action by generating text such as click(12), although most of that text encodes a choice among a few action types and a few…
LLM-Based Test Generation: Information Sources, Generation Strategies, and Quality Evidence
Yunhao Liang, Chengguang Gan, Ruixuan Ying +3
Large language models are increasingly used to generate test scenarios, executable test suites, assertions, interaction sequences, and fuzzing infrastructure. These artifacts serve…
Certified Selective Automation of LLM Agent Evaluation
Chengguang Gan, Yunhao Liang, Qinghao Zhang +1
Evaluating LLM agents still ends with a human reading trajectories, because automatic judges carry no guarantee on how often they are wrong. We ask the operational question: what f…
Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents
Chengguang Gan, Yunhao Liang, QingHao Zhang +1
Web agents are usually evaluated in live environments, where environment state and judge models drift between runs, so the same checkpoint rarely reproduces the same score, making…
Test-Driven Approaches to Software Engineering with Large Language Models: A Survey of Phases, Tasks, and Agent Skills
Yunhao Liang, Chengguang Gan, Ruixuan Ying +3
Tests increasingly participate in the decisions made by large language models and software engineering agents. They specify intended behavior, guide program construction and repair…
Correct Tests Are Not Enough: Measuring and Training Oracle Conversion in Specification-Based Test Generation
Yunhao Liang, Chengguang Gan, Ruixuan Ying +3
Generating tests from a natural-language specification requires both an input that exposes faulty behavior and a correct expected output. These requirements need not improve togeth…
How Output Format Confounds Data Quality and Capability in Instruction Tuning
Chengguang Gan, Hanjun Wei, Yunhao Liang +3
Instruction-tuning data are judged by quality metrics, and tuned models are judged by benchmarks, but both judgments pass through an output interface: the surface format in which a…
Joint Training Is Not Enough: Conditioned Cross-Granularity Training for Multimodal Document Understanding
Chengguang Gan, Yunhao Liang, Hanjun Wei +2
The Mutual Reinforcement Effect (MRE) asks whether a fine, span-level and a coarse, document-level task help each other when one model handles both. We test it in multimodal docume…
Auditing and Decomposing Feedback-Driven Evolution in LLM Test Generation under the Oracle Problem
Yunhao Liang, Chengguang Gan, Ruixuan Ying +3
Execution feedback is often treated as a self-verifying signal for improving LLM-generated tests. However, when generated inputs are executed on a single accepted program and its o…
Security Tests as Executable Specifications for LLM Code Generation: Benefits, Trade-offs, and Coverage Limits
Yunhao Liang, Chengguang Gan, Ruixuan Ying +3
Large language models (LLMs) can generate functionally useful code that remains vulnerable, while security-focused interventions may break intended behavior. We investigate securit…
MAG: A Web-Agent Benchmark and Harness for Multimodal Action and Guide Generation
Chengguang Gan, Hanjun Wei, Yunhao Liang +3
Digital Adoption Platforms (DAPs) are embedded overlays widely used on web systems to guide users through operations inside a page, helping them get started with unfamiliar interfa…
A Learning-Rate-Gated Failure of GRPO in a Small Language and Vision-Language Model Web Agent: A Controlled Null and Its Mechanism
Chengguang Gan, Zhixi Cai, Yunhao Liang +3
Reinforcement learning with verifiable rewards, and Group Relative Policy Optimization (GRPO) in particular, is now run routinely on a supervised checkpoint in the hope of producin…
Do Code Language Models Follow Tests? Paired Interventions on Program Behavior
Yunhao Liang, Chengguang Gan, Ruixuan Ying +3
Visible tests specify concrete program behavior, but an improvement in benchmark accuracy does not establish that a model follows the rule expressed by those tests. We study test u…
Scaling Test-Driven Code Generation from Functions to Classes: An Empirical Study
Yunhao Liang, Ruixuan Ying, Shiwen Ni +1
Test-driven development (TDD) has been adopted to improve Large Language Model (LLM)-based code generation by using tests as executable specifications. However, existing TDD-style…
GuideWeb: A Benchmark for Automatic In-App Guide Generation on Real-World Web UIs
Chengguang Gan, Yoshihiro Tsujii, Yunhao Liang +3
Digital Adoption Platform (DAP) provide web-based overlays that deliver operation guidance and contextual hints to help users navigate complex websites. Although modern DAP tools e…
Earley-Driven Dynamic Pruning for Efficient Structured Decoding
Xintong Sun, Chi Wei, Minghao Tian +1
Large Language Models (LLMs) have shown remarkable capabilities, yet ensuring their outputs conform to strict structural or grammatical constraints remains challenging, which is cr…
ScaleLong: A Multi-Timescale Benchmark for Long Video Understanding
David Ma, Huaqing Yuan, Xingjian Wang +16
Although long-video understanding demands that models capture hierarchical temporal information -- from clip (seconds) and shot (tens of seconds) to event (minutes) and story (hour…
IV-Bench: A Benchmark for Image-Grounded Video Perception and Reasoning in Multimodal LLMs
David Ma, Yuanxing Zhang, Jincheng Ren +17
Existing evaluation frameworks for Multimodal Large Language Models (MLLMs) primarily focus on image reasoning or general video understanding tasks, largely overlooking the signifi…
M-MRE: Extending the Mutual Reinforcement Effect to Multimodal Information Extraction
Chengguang Gan, Zhixi Cai, Yanbin Wei +3
Mutual Reinforcement Effect (MRE) is an emerging subfield at the intersection of information extraction and model interpretability. MRE aims to leverage the mutual understanding be…
A Multilingual Dataset and Empirical Validation for the Mutual Reinforcement Effect in Information Extraction
Chengguang Gan, Sunbowen Lee, Qingyu Yin +9
The Mutual Reinforcement Effect (MRE) describes a phenomenon in information extraction where word-level and sentence-level tasks can mutually improve each other when jointly modele…