Abstract
Recommendation systems have been essential for both user experience andplatform efficiency by alleviating information overload and supportingdecision-making. Traditional methods, i.e., content-based filtering,collaborative filtering, and deep learning, have achieved impressive results inrecommendation systems. However, the cold-start and sparse-data scenarios arestill challenging to deal with. Existing solutions either generatepseudo-interaction sequence, which often introduces redundant or noisy signals,or rely heavily on semantic similarity, overlooking dynamic shifts in usermotivation. To address these limitations, this paper proposes a novelrecommendation framework, termed M-$LLM^3$REC, which leverages large languagemodels for deep motivational signal extraction from limited user interactions.M-$LLM^3$REC comprises three integrated modules: the Motivation-OrientedProfile Extractor (MOPE), Motivation-Oriented Trait Encoder (MOTE), andMotivational Alignment Recommender (MAR). By emphasizing motivation-drivensemantic modeling, M-$LLM^3$REC demonstrates robust, personalized, andgeneralizable recommendations, particularly boosting performance in cold-startsituations in comparison with the state-of-the-art frameworks.