Do Xuan Long, Duy Dinh, Ngoc-Hai Nguyen, Kenji Kawaguchi, Nancy F. Chen, Shafiq Joty, Min-Yen Kan. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Reasoning and predicting human opinions with large language models (LLMs) is essential yet challenging. Current methods employ role-playing with personae but face two major issues: LLMs are sensitive to even a single irrelevant persona, skewing predictions by up to 30%; and LLMs fail to reason strategically over personae. We propose Chain-of-Opinion (COO), a simple four-step solution modeling which and how to reason with personae, inspired by the Value–Belief–Norm (VBN) theory. COO differentiates between explicit personae (demographics and ideology) and implicit personae (historical opinions), involves: (1) filtering irrelevant attributes from explicit personae; (2) ranking implicit personae into a preferential list for selecting top-k; (3) applying novel VBN reasoning to extract user environmental and personal value, belief, and norm variables for accurate and reliable predictions; and (4) iterating VBN reasoning with progressively larger lists of implicit personae to handle potential persona insufficiency. COO efficiently achieves new state-of-the-art opinion prediction via prompting with only 5 inference calls, improving prior techniques by up to 4%. Notably, fine-tuning LMs with COO’s data results in significantly better opinion-aligned models, by up to 23%.

Do Xuan Long, Ngoc-Hai Nguyen, Tiviatis Sim, Hieu Dao, Shafiq Joty, Kenji Kawaguchi, Nancy F. Chen, Min-Yen Kan. 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.

Xuan Long Do, Yiran Zhao, Hannah Brown, Yuxi Xie, James Xu Zhao, Nancy F. Chen, Kenji Kawaguchi, Michael Shieh, Junxian He. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

We present Multi-expert Prompting 1 , a novel enhancement of ExpertPrompting (Xu et al., 2023), designed to improve the large language model (LLM) generation.Specifically, it guides an LLM to fulfill an input instruction by simulating multiple experts, aggregating their responses, and selecting the best among individual and aggregated responses.This process is performed in a single chain of thoughts through our seven carefully designed subtasks derived from the Nominal Group Technique (Ven and Delbecq, 1974), a well-established decision-making framework.Our evaluations demonstrate that Multi-expert Prompting significantly outperforms ExpertPrompting and comparable baselines in enhancing the truthfulness, factuality, informativeness, and usefulness of responses while reducing toxicity and hurtfulness.It further achieves state-ofthe-art truthfulness by outperforming the best baseline by 8.69% with ChatGPT.Multi-expert Prompting is efficient, explainable, and highly adaptable to diverse scenarios, eliminating the need for manual prompt construction. * Equal contribution.1 Our codes and data will be made publicly available here.

Nhat M. Hoang, Xuan Long Do, Duc Anh Do, Duc Anh Vu, Luu Anh Tuan. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2024.

Safety of Large Language Models (LLMs) has become a central issue given their rapid progress and wide applications. Greedy Coordinate Gradient (GCG) is shown to be effective in constructing prompts containing adversarial suffixes to break the presumingly safe LLMs, but the optimization of GCG is time-consuming and limits its practicality. To reduce the time cost of GCG and enable more comprehensive studies of LLM safety, in this work, we study a new algorithm called $\texttt{Probe sampling}$ to accelerate the GCG algorithm. At the core of the algorithm is a mechanism that dynamically determines how similar a smaller draft model's predictions are to the target model's predictions for prompt candidates. When the target model is similar to the draft model, we rely heavily on the draft model to filter out a large number of potential prompt candidates to reduce the computation time. Probe sampling achieves up to $5.6$ times speedup using Llama2-7b-chat and leads to equal or improved attack success rate (ASR) on the AdvBench. Furthermore, probe sampling is also able to accelerate other prompt optimization techniques and adversarial attack methods, leading to acceleration of $1.8\times$ for AutoPrompt, $2.4\times$ for APE and $2.4\times$ for AutoDAN.

Xuan Long Do, Bowei Zou, Shafiq Joty, Tran Tai, Liangming Pan, Nancy Chen, Ai Ti Aw. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Charts are widely used for data analysis, providing visual representations and insights into complex data.To facilitate chart-based data analysis using natural language, several downstream tasks have been introduced recently such as chart question answering and chart summarization.However, existing methods for these tasks often rely on pretraining on language or vision-language tasks, neglecting the explicit modeling of chart structures (e.g., how chart elements are related to each other).To address this, we first build a large corpus of charts covering diverse topics and visual styles.We then present UniChart, a pretrained model for chart comprehension and reasoning.UniChart encodes the relevant text, data, and visual elements of charts and then uses a chart-grounded text decoder for text generation.We propose several chart-specific pretraining tasks that include: (i) low-level tasks to extract the visual elements (e.g., bars, lines) and data from charts, and (ii) high-level tasks to acquire chart understanding and reasoning skills.Our experiments demonstrate that pretraining UniChart on a large corpus with chart-specific objectives, followed by fine-tuning, yields state-of-the-art performance on four downstream tasks.Moreover, our model exhibits superior generalizability to unseen chart corpus, surpassing previous approaches that lack chart-specific objectives and utilize limited chart resources.

Conversational question generation (CQG) serves as a vital task for machines to assist humans, such as interactive reading comprehension, through conversations. Compared to traditional single-turn question generation (SQG), CQG is more challenging in the sense that the generated question is required not only to be meaningful, but also to align with the provided conversation. Previous studies mainly focus on how to model the flow and alignment of the conversation, but do not thoroughly study which parts of the context and history are necessary for the model. We believe that shortening the context and history is crucial as it can help the model to optimise more on the conversational alignment property. To this end, we propose CoHS-CQG, a two-stage CQG framework, which adopts a novel CoHS module to shorten the context and history of the input. In particular, it selects the top-p sentences and history turns by calculating the relevance scores of them. Our model achieves state-of-the-art performances on CoQA in both the answer-aware and answer-unaware settings.

Charts are very popular to analyze data and convey important insights. People often analyze visualizations to answer open-ended questions that require explanatory answers. Answering such questions are often difficult and time-consuming as it requires a lot of cognitive and perceptual efforts. To address this challenge, we introduce a new task called OpenCQA, where the goal is to answer an open-ended question about a chart with descriptive texts. We present the annotation process and an in-depth analysis of our dataset. We implement and evaluate a set of baselines under three practical settings. In the first setting, a chart and the accompanying article is provided as input to the model. The second setting provides only the relevant paragraph(s) to the chart instead of the entire article, whereas the third setting requires the model to generate an answer solely based on the chart. Our analysis of the results show that the top performing models generally produce fluent and coherent text while they struggle to perform complex logical and arithmetic reasoning.