Elamur Research⌕
☰
home / papers / 2026 / IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters

IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters

Paper · 2026 · Hongwei Zheng, Weiqi Wu, Zhengjia Wang et al.

Summary

The paper addresses a gap in proactive conversational agents. Prior work makes assistants more active only after a dialogue has started. Before that, many users open the app with a vague need, never type anything and leave. Production logs show about 20% of users do this. The authors define this as the task of Conversation Starter Generation: suggesting a few personalized opening questions with no explicit intent to work from.

Their framework, IceBreaker from ByteDance, works in two steps, modeled on how people break the ice:

  1. Resonance-Aware Interest Distillation (RID) filters a user's past session summaries down to a few "trigger interests." It uses a contrastively trained resonance scorer and thresholds that depend on how active the user is.
  2. Interaction-Oriented Starter Generation (ISG) writes a short, varied list of starters from those interests. It is first instruction-tuned on teacher-LLM data, then refined with list-wise preference optimization that balances utility and diversity.

In month-long online A/B tests on a very large commercial assistant, IceBreaker increased engagement and click-through and is now in production. This shows that the cold-start moment is a practical lever for retention.

Key points

  • Online results: user active days rose by +1.84‰ and click-through rate by +94.25‰. Gains also appeared in conversation start rate (CSR) and downstream engagement metrics (Active, Avg.S.).
  • RID scorer: cosine similarity between user and summary embeddings. It is trained with a contrastive loss, using "interest revisits" in later sessions as positives and both within-user and cross-user negatives.
  • Activity-aware gating: users are grouped by activity level, and each group gets its own threshold, which rises for more active users.
  • ISG training:
    • Supervised Interest-Expanded Instruction Tuning (SIT) on a teacher-distilled corpus.
    • Personalized Multi-Dimensional Alignment (PMA), which uses a hybrid utility-plus-diversity beam list search to build hierarchical preference pairs, refreshed iteratively.
  • Base model and analysis: the Doubao-based variant performed best offline and was the one deployed. Topic analysis shows RID removes functional noise (e.g., translation requests) and produces broader, less head-dominated topic coverage than the earlier retrieval-based starters.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksPsychology · 1,649 worksSocial Psychology · 330 works

🧭 Topics 1Research subjects and languages the work is about.Dialogue systems · 654 works

Cite

Cite as Hongwei Zheng, Weiqi Wu, Zhengjia Wang, Guanyu Jiang, Haoming Li, Tianyu Wu, Yongchun Zhu, Jingwu Chen and Feng Zhang. 2026. IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-industry.16

@inproceedings{zheng2026icebreaker,
  title = {{IceBreaker for Conversational Agents: Breaking the First-Message Barrier with Personalized Starters}},
  author = {Hongwei Zheng, Weiqi Wu, Zhengjia Wang, Guanyu Jiang, Haoming Li, Tianyu Wu, Yongchun Zhu, Jingwu Chen and Feng Zhang},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-industry.16},
  url = {https://doi.org/10.18653/v1/2026.acl-industry.16}
}

References

26 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Chinmaya Andukuri, Jan-Philipp Fränken, Tobias Gerstenberg, and Noah Goodman. 2024. STar-GATE: Teaching language models to ask clarifying questions. In First Conference on Language Modeling. find
  2. Jiaxin Deng, Shiyao Wang, Kuo Cai, Lejian Ren, Qigen Hu, Weifeng Ding, Qiang Luo, and Guorui Zhou. 2025a. Onerec: Unifying retrieve and rank with generative recommender and iterative preference alignment. arXiv preprint arXiv:2502.18965. — [arXiv:2502.18965](https://arxiv.org/abs/2502.18965) find
  3. Yang Deng, Lizi Liao, Liang Chen, Hongru Wang, Wenqiang Lei, and Tat-Seng Chua. 2023. Prompting and evaluating large language models for proactive dialogues: Clarification, target-guided, and non-collaboration. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10602–106…
  4. Yang Deng, Lizi Liao, Wenqiang Lei, Grace Hui Yang, Wai Lam, and Tat-Seng Chua. 2025b. Proactive conversational ai: A comprehensive survey of advancements and opportunities. ACM Transactions on Information Systems, 43(3):1–45. find
  5. Xian Guo, Ben Chen, Siyuan Wang, Ying Yang, Chenyi Lei, Yuqing Ding, and Han Li. 2025. Onesug: The unified end-to-end generative framework for e-commerce query suggestion. arXiv preprint arXiv:2506.06913. — [arXiv:2506.06913](https://arxiv.org/abs/2506.06913) find
  6. Muhammad Usman Hadi, Rizwan Qureshi, Abbas Shah, Muhammad Irfan, Anas Zafar, Muhammad Bilal Shaikh, Naveed Akhtar, Jia Wu, Seyedali Mirjalili, and 1 others. 2023. A survey on large language models: Applications, challenges, limitations, and practical usage. Authorea Preprints. find
  7. Enkelejda Kasneci, Kathrin Seßler, Stefan Küchemann, Maria Bannert, Daryna Dementieva, Frank Fischer, Urs Gasser, Georg Groh, Stephan Günnemann, Eyke Hüllermeier, and 1 others. 2023. Chatgpt for good? on opportunities and challenges of large language models for education. Learning and individual dif…
  8. Han Li, Renwen Zhang, Yi-Chieh Lee, Robert E Kraut, and David C Mohr. 2023. Systematic review and meta-analysis of ai-based conversational agents for promoting mental health and well-being. NPJ Digital Medicine, 6(1):236. find
  9. Hao Li, Chenghao Yang, An Zhang, Yang Deng, Xiang Wang, and Tat-Seng Chua. 2025a. Hello again! LLM-powered personalized agent for long-term dialogue. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technolo…
  10. Xiaoyu Li, Xiao Li, Li Gao, Yiding Liu, Xiaoyang Wang, Shuaiqiang Wang, Junfeng Wang, and Dawei Yin. 2025b. Proactive guidance of multi-turn conversation in industrial search. arXiv preprint arXiv:2505.24251. find
  11. Jiayi Liao, Xiangnan He, Ruobing Xie, Jiancan Wu, Yancheng Yuan, Xingwu Sun, Zhanhui Kang, and Xiang Wang. 2024. Rosepo: Aligning llm-based recommenders with human values. ArXiv, abs/2410.12519. — [arXiv:2410.12519](https://arxiv.org/abs/2410.12519) find
  12. Na Liu, Liangyu Chen, Xiaoyu Tian, Wei Zou, Kaijiang Chen, and Ming Cui. 2024. From llm to conversational agent: A memory enhanced architecture with fine-tuning of large language models. arXiv preprint arXiv:2401.02777. — [arXiv:2401.02777](https://arxiv.org/abs/2401.02777) find
  13. Chen Luo, Dimitri Papadimitriou, Hariharan Muralidharan, Dhineshkumar Ramasubbu, Aakash Kolekar, Wenju Xu, Cong Xu, Anirudh Srinivasan, Mukesh Jain, and Qi He. 2025. Language model alignment for conversational shopping at amazon. In Proceedings of the 48th International ACM SIGIR Conference on Resea… find
  14. Don Norman. 2013. The design of everyday things: Revised and expanded edition. Basic books. find
  15. Weicong Qin, Yi Xu, Weijie Yu, Chenglei Shen, Ming He, Jianping Fan, Xiao Zhang, and Jun Xu. 2025. MAPS: Motivation-aware personalized search via LLM-driven consultation alignment. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pag… find
  16. Yujia Qin, Shengding Hu, Yankai Lin, Weize Chen, Ning Ding, Ganqu Cui, Zheni Zeng, Xuanhe Zhou, Yufei Huang, Chaojun Xiao, and 1 others. 2024. Tool learning with foundation models. ACM Computing Surveys, 57(4):1–40.
  17. Traian Rebedea, Makesh Sreedhar, Shaona Ghosh, Jiaqi Zeng, and Christopher Parisien. 2024. Canttalkaboutthis: Aligning language models to stay on topic in dialogues. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 12232– 12252. find
  18. Chao Sun, Yaobo Liang, Yaming Yang, Shilin Xu, Tianmeng Yang, and Yunhai Tong. 2024. Rlrf4rec: Reinforcement learning from recsys feedback for enhanced recommendation reranking. ArXiv, abs/2410.05939. — [arXiv:2410.05939](https://arxiv.org/abs/2410.05939) find
  19. Jianling Wang, Yifan Liu, Yinghao Sun, Xuejian Ma, Yueqi Wang, He Ma, Zhengyang Su, Minmin Chen, Mingyan Gao, Onkar Dalal, Ed H. Chi, Lichan Hong, Ningren Han, and Haokai Lu. 2025. User feedback alignment for LLM-powered exploration in large-scale recommendation systems. In Proceedings of the 63rd A… find
  20. Yufei Wang, Wanjun Zhong, Liangyou Li, Fei Mi, Xingshan Zeng, Wenyong Huang, Lifeng Shang, Xin Jiang, and Qun Liu. 2023. Aligning large language models with human: A survey. Preprint, arXiv:2307.12966.
  21. Shujin Wu, Yi R. Fung, Cheng Qian, Jeonghwan Kim, Dilek Hakkani-Tur, and Heng Ji. 2025. Aligning LLMs with individual preferences via interaction. In Proceedings of the 31st International Conference on Computational Linguistics, pages 7648–7662, Abu Dhabi, UAE. Association for Computational Linguist…
  22. Zhongheng Yang, Aijia Sun, Yushang Zhao, Yinuo Yang, Dannier Li, and Chengrui Zhou. 2025. Rlhf fine-tuning of llms for alignment with implicit user feedback in conversational recommenders. Preprint, arXiv:2508.05289. — [arXiv:2508.05289](https://arxiv.org/abs/2508.05289) find
  23. Junhao Yin, Haolin Wang, Peng Bao, Ju Xu, and Yongliang Wang. 2025. From clicks to preference: A multi-stage alignment framework for generative query suggestion in conversational system. Preprint, arXiv:2508.15811. — [arXiv:2508.15811](https://arxiv.org/abs/2508.15811) find
  24. J Diego Zamfirescu-Pereira, Richmond Y Wong, Bjoern Hartmann, and Qian Yang. 2023. Why johnny can’t prompt: how non-ai experts try (and fail) to design llm prompts. In Proceedings of the 2023 CHI conference on human factors in computing systems, pages 1–21. find
  25. Jinhui Zhu, Xiangfeng Luo, Xin Yao, and Xiao Wei. 2026. Lrsa: Llm-recsys alignment for time-specific next poi recommendation. Information Processing & Management, 63(2, Part B):104434. find
  26. Yutao Zhu, Huaying Yuan, Shuting Wang, Jiongnan Liu, Wenhan Liu, Chenlong Deng, Haonan Chen, Zheng Liu, Zhicheng Dou, and Ji-Rong Wen. 2025. Large language models for information retrieval: A survey. ACM Transactions on Information Systems, 44(1):1–54.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zheng-2026-icebreaker-conversational-agents-breaking-first/