Polite language production involves balancing strategic success with local convention, but these pressures are difficult to separate in natural interaction. We therefore examine them in an artificial-language request game, where one marker increased request acceptance and another appeared more often in other speakers’ requests. Humans and LLMs completed the same task under goals to maximize acceptance, match local usage, or produce a natural and appropriate request. For both humans and LLMs, a policy-mixture model combining payoff- and exposure-sensitive policies provided the best overall predictive fit. However, their baseline behavior differed. Human baseline behavior fell between the strategy and convention conditions, suggesting sensitivity to both payoff and local usage. LLMs shifted toward local usage under the convention goal, but their baselines remained closer to the strategy condition. These results show that human and LLM behavior is sensitive to both payoff and local usage in polite production, but that LLM baselines are more payoff-oriented. Code, data, and materials are all available at https://github.com/CHENTianfang/emnlp2026-politeness.

Collecting and constructing human-annotated corpora for training conversational question-answering (CQA) models has recently been shown to be inefficient and costly. To solve this problem, previous works have proposed training QA models with automatically generated QA data. In this work, we extend earlier studies on QA synthesis, and propose an efficient QA data generation algorithm under conversational settings. Our model recognizes potential dialogue topics, generates corresponding questions, and extracts answers from grounding passages. To improve the quality of generated QAs and downstream self-training of CQA models, we propose dropout and agreement-based QA selection methods. We conduct experiments on both data augmentation and domain adaptation settings. Experiments on the QuAC and Doc2Dial tasks show that the proposed method can significantly improve the quality of generated QA data, and also improves the accuracy of self-trained CQA models based on the constructed training corpora.

MultiDoc2Dial presents an important challenge on modeling dialogues grounded with multiple documents. This paper proposes a pipeline system of “retrieve, re-rank, and generate”, where each component is individually optimized. This enables the passage re-ranker and response generator to fully exploit training with ground-truth data. Furthermore, we use a deep cross-encoder trained with localized hard negative passages from the retriever. For the response generator, we use grounding span prediction as an auxiliary task to be jointly trained with the main task of response generation. We also adopt a passage dropout and regularization technique to improve response generation performance. Experimental results indicate that the system clearly surpasses the competitive baseline and our team CPII-NLP ranked 1st among the public submissions on ALL four leaderboards based on the sum of F1, SacreBLEU, METEOR and RougeL scores.