Yunhao Liang

Author directory

2026

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 modeled. While prior work has reported MRE in Japanese, its generality across languages and task settings has not been empirically validated, largely due to the lack of multilingual MRE datasets. To address this limitation, we introduce the Multilingual MRE Mix dataset (MMM), which consists of 21 sub-datasets covering English, Japanese, and Chinese. We propose an LLM-assisted dataset translation and alignment framework that significantly reduces manual annotation effort while preserving the structural requirements of MRE tasks. Building on MMM, we adopt a unified input-output framework to train an open-domain information extraction model and conduct extensive empirical studies, including full fine-tuning ablations and the construction of knowledgeable verbalizers based on MRE-mix data. Experimental results show that 76 percent of the MMM sub-datasets consistently exhibit the Mutual Reinforcement Effect across languages. These findings provide systematic empirical validation of MRE in multilingual settings and demonstrate its practical value for information extraction.
Digital Adoption Platforms (DAPs) provide web-based overlays that deliver operational guidance and contextual hints to help users navigate complex websites. Although modern DAP tools enable non-experts to author such guidance, maintaining guides remains labor-intensive because website layouts and functionalities evolve continuously. In this work, we introduce GuideWeb, a human-verified benchmark for automatic in-app guide proposal on real-world web UIs. GuideWeb formulates the task as selecting guide target elements grounded in the webpage and generating concise guide text aligned with user intent. We make the target audience and guide-worthiness criteria explicit, and construct annotations through LLM-assisted candidate discovery followed by human verification, including deletion, rewriting, grounding correction, and addition of missed targets. We also propose an evaluation suite that jointly measures guide target selection, DOM grounding, and generated intent/guide-text quality. Experiments show that our proposed GuideWeb Agent achieves 30.79% F1 in guide target prediction, while obtaining BLEU scores of 44.94 for intent generation and 21.34 for guide-text generation. Existing baselines perform substantially worse, highlighting that reliable automatic guide proposal remains challenging on real-world web UIs.