Effectively providing emotional support is a critical yet intricate interpersonal skill. Supporters often lack accessible and practical training opportunities to develop this competency. To address this gap, we introduce ESR-Coach, a Large Language Model (LLM)-based coaching system designed to train individuals in emotionally supportive communication. ESR-Coach leverages multiple AI agents to generate practice scenarios, demonstrate reference responses, and provide assessments on user practice replies. We evaluate the proficiency of our system on these three tasks, demonstrating high-fidelity case generation, helpful exemplary responses, and valid response assessments. In our user study (N=20), ESR-Coach helped participants achieve an average improvement of 17% in response helpfulness. After training, participants also employed more diverse and effective strategies. We further discuss the social intelligence of LLMs and their potential to foster humans’ interpersonal skills in real-world scenarios.

Vision Language Models (VLMs) often struggle with chart understanding tasks, particularly in accurate chart description and complex reasoning. Synthetic data generation is a promising solution, while usually facing the challenge of noise labels. To address this challenge, we first introduce a chart synthesis pipeline that generates aligned chart-question-answer triplets through code generation and execution, ensuring the reliability of synthetic data without human intervention. Furthermore, inspired by test-time scaling that increases inference budget and thereby improves performance, we design a candidate-conditioned answering process. The VLM first generates multiple responses per query, and then synthesizes the final answer by contextualizing these candidates. Experiments demonstrate significant improvements, with up to 15.50 points accuracy gain over the initial VLM, in a fully self-improving paradigm without either human-labeled data or external models.

Gongyao Jiang, Xinran Shi, Qiong Luo. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

Zhuang Chen, Jincenzi Wu, Jinfeng Zhou, Bosi Wen, Guanqun Bi, Gongyao Jiang, Yaru Cao, Mengting Hu, Yunghwei Lai, Zexuan Xiong, Minlie Huang. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Abstract Dialogue-level dependency parsing, despite its growing academic interest, often encounters underperformance issues due to resource shortages. A potential solution to this challenge is data augmentation. In recent years, large language models (LLMs) have demonstrated strong capabilities in generation, which can facilitate data augmentation greatly. In this study, we focus on Chinese dialogue-level dependency parsing, presenting three simple and effective strategies with LLM to augment the original training instances, namely word-level, syntax-level, and discourse-level augmentations, respectively. These strategies enable LLMs to either preserve or modify dependency structures, thereby assuring accuracy while increasing the diversity of instances at different levels. We conduct experiments on the benchmark dataset released by Jiang et al. (2023) to validate our approach. Results show that our method can greatly boost the parsing performance in various settings, particularly in dependencies among elementary discourse units. Lastly, we provide in-depth analysis to show the key points of our data augmentation strategies.

In many studies, the graph convolution neural networks were used to solve different natural language processing (NLP) problems. However, few researches employ graph convolutional network for text classification, especially for short text classification. In this work, a special text graph of the short-text corpus is created, and then a short-text graph convolutional network (STGCN) is developed. Specifically, different topic models for short text are employed, and a short text short-text graph based on the word co-occurrence, document word relations, and text topic information, is developed. The word and sentence representations generated by the STGCN are considered as the classification feature. In addition, a pre-trained word vector obtained by the BERTs hidden layer is employed, which greatly improves the classification effect of our model. The experimental results show that our model outperforms the state-of-the-art models on multiple short text datasets.