{"id":"implicit-user-feedback","title":"User Feedback in Human-LLM Dialogues: A Lens to Understand Users But Noisy as a Learning Signal","authors":[{"given":"Yuhan","family":"Liu"},{"given":"Michael J. Q.","family":"Zhang"},{"given":"Eunsol","family":"Choi"}],"first_public":"2025-07-30","date_note":"First arXiv version.","date_source":"https://arxiv.org/abs/2507.23158","year":2025,"venue":"EMNLP 2025","paper_url":"https://aclanthology.org/2025.emnlp-main.133/","arxiv":"2507.23158","category":"behavior","relationship":"uses","dataset_version":"WildChat and LMSYS-Chat-1M, including dense manual annotations on 34 WildChat conversations and separate 1,000-conversation samples for feedback-conditioned response experiments.","summary":"Examines the frequency, meaning, and usefulness of implicit feedback in real conversations. It separates feedback polarity from what users actually request, and tests whether regenerating responses with that additional context improves model training.","finding":"Feedback semantics help on MTBench but do not consistently improve the longer, more complex WildBench tasks; positive feedback alone can be a noisy training signal.","evidence_url":"https://arxiv.org/html/2507.23158v2","evidence_note":"Sections 3–4: manual labels and user behavior. Sections 5–7: regeneration and training. Table 1: the 34-conversation WildChat annotation sample.","bibtex_key":"implicituserfeedback2025","record_url":"https://yuntiandeng.com/wildchat/papers/implicit-user-feedback/","relationship_label":"Uses WildChat","citation_title":"User Feedback in Human-LLM Dialogues: A Lens to Understand Users But Noisy as a Learning Signal","citation_type":"paper-conference","bibtex_type":"inproceedings","citation_url":"https://aclanthology.org/2025.emnlp-main.133/","citation_year":2025,"citation_note":"EMNLP 2025 proceedings paper.","doi":"10.18653/v1/2025.emnlp-main.133","code_url":"https://github.com/lyh6560new/implicit-user-feedback","container_title":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","pages":"2666–2681","publisher":"Association for Computational Linguistics"}