{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T02:54:49Z","timestamp":1773802489085,"version":"3.50.1"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"18","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AAAI"],"abstract":"<jats:p>Large Language Models (LLMs) have recently emerged as powerful reasoning engines in recommender systems, generating natural-language explanations that foster user engagement.\nHowever, their recommendation performance remains limited, as they lack exposure to collaborative user-item interaction patterns.\nIn contrast, collaborative filtering (CF) models achieve strong performance by learning from these behavioral patterns at scale.\nTo unify the strengths of both paradigms, we propose TWiCE-Rec (Think Wise, Collaborate Effectively), a rationale-aware LLM-based recommender that incorporates collaborative user-item interactions.\nIn the first stage, we construct a rationale dataset by applying in-context learning with self-annotated curation. \nA state-of-the-art LLM is guided to generate persuasive rationales that explain the causal relationship between the user\u2019s interaction sequence and the ground-truth next item, resulting in a curated post-hoc training dataset.\nIn the second stage, we perform multi-task instruction-tuned adaptation\u2014based on the rationale-augmented training dataset\u2014comprising item description generation and both non-reasoning and reasoning-based sequential recommendation, to equip the LLM with the ability to generate rationales that reflect how user preferences align with item characteristics.\nFinally, we aim to enhance the LLM\u2019s recommendation performance by incorporating user-item interaction patterns derived from the CF-Rec model.\nTo achieve this, we propose a confidence-weighted reinforcement learning strategy that adjusts rewards in proportion to both the LLM\u2019s prediction alignment with the ground-truth and the confidence from the pretrained CF-Rec model.\nOur method outperforms both CF- and LLM-Rec models on Amazon datasets in terms of recommendation performance and rationale quality. \nIn an online A\/B test, it achieved about 8% higher click-through rate than existing models, demonstrating practical value.<\/jats:p>","DOI":"10.1609\/aaai.v40i18.38590","type":"journal-article","created":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T00:36:26Z","timestamp":1773794186000},"page":"15609-15616","source":"Crossref","is-referenced-by-count":0,"title":["Think Wise, Collaborate Effectively: A Rationale-Aware LLM-Based Recommender with Reinforcement Learning from Collaborative Signals"],"prefix":"10.1609","volume":"40","author":[{"given":"Chung","family":"Park","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Taesan","family":"Kim","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hyeongjun","family":"Yun","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Dongjoon","family":"Hong","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Junui","family":"Hong","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Kijung","family":"Park","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"MinCheol","family":"Cho","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Min Sung","family":"Choi","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jihwan","family":"Seok","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jaegul","family":"Choo","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"9382","published-online":{"date-parts":[[2026,3,14]]},"container-title":["Proceedings of the AAAI Conference on Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/38590\/42552","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/download\/38590\/42552","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,18]],"date-time":"2026-03-18T00:36:27Z","timestamp":1773794187000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/view\/38590"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,14]]},"references-count":0,"journal-issue":{"issue":"18","published-online":{"date-parts":[[2026,3,17]]}},"URL":"https:\/\/doi.org\/10.1609\/aaai.v40i18.38590","relation":{},"ISSN":["2374-3468","2159-5399"],"issn-type":[{"value":"2374-3468","type":"electronic"},{"value":"2159-5399","type":"print"}],"subject":[],"published":{"date-parts":[[2026,3,14]]}}}