{"id":"wildreward","title":"WildReward: Learning Reward Models from In-the-Wild Human Interactions","authors":[{"given":"Hao","family":"Peng"},{"given":"Yunjia","family":"Qi"},{"given":"Xiaozhi","family":"Wang"},{"given":"Zijun","family":"Yao"},{"given":"Lei","family":"Hou"},{"given":"Juanzi","family":"Li"}],"first_public":"2026-02-09","date_note":"First arXiv version.","date_source":"https://arxiv.org/abs/2602.08829","year":2026,"venue":"ACL 2026","paper_url":"https://aclanthology.org/2026.acl-long.864/","arxiv":"2602.08829","category":"training","relationship":"extends","dataset_version":"Filtered English and Chinese conversations from WildChat-4.8M, yielding approximately 186,000 feedback instances for the WildFB training data.","summary":"Trains reward models directly on implicit user feedback using ordinal regression. Its pipeline classifies feedback, filters ambiguous signals, and validates refusal cases before training; experiments examine calibration, cross-prompt score consistency, user diversity, and online DPO.","finding":"Broader user diversity improves reward modeling. WildReward models also improve online DPO results over the reward-model baseline tested in the paper.","evidence_url":"https://arxiv.org/html/2602.08829v1","evidence_note":"Section 2: mining and training. Section 4: score consistency and user diversity. Appendix A: WildChat-4.8M language and quality filters.","bibtex_key":"wildreward2026","record_url":"https://yuntiandeng.com/wildchat/papers/wildreward/","relationship_label":"Extends WildChat","citation_title":"WildReward: Learning Reward Models from In-the-Wild Human Interactions","citation_type":"paper-conference","bibtex_type":"inproceedings","citation_url":"https://aclanthology.org/2026.acl-long.864/","citation_year":2026,"citation_note":"ACL 2026 proceedings paper.","doi":"10.18653/v1/2026.acl-long.864","code_url":"https://github.com/THU-KEG/WildReward","container_title":"Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","pages":"18932–18948","publisher":"Association for Computational Linguistics"}