TwiSTAR:Think Fast, Think Slow, Then Act,Generative Recommendation with Adaptive Reasoning

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Cao, Shiteng, Jiang, Kaian, Gong, Yunlong, Li, Zhiheng
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917483689017344
author Cao, Shiteng
Jiang, Kaian
Gong, Yunlong
Li, Zhiheng
author_facet Cao, Shiteng
Jiang, Kaian
Gong, Yunlong
Li, Zhiheng
contents Generative recommendation with Semantic IDs (SIDs) has emerged as a promising paradigm, yet existing methods apply a fixed inference strategy, either fast direct generation or slow chain-of-thought reasoning, uniformly across all user histories. This approach creates a trade-off: fast recommendation model produces suboptimal accuracy on hard samples, while always invoking slow reasoning incurs prohibitive latency and wastes computation on easy cases. To address this, we propose Think Fast, Think Slow, Then Act, a framework that learns to adaptively allocate reasoning effort per user sequence. Our system equips an LLM with three complementary tools: a fast SID-based retriever, a lightweight candidate ranker, and a slow reasoning model that generates explicit rationales before recommending. Crucially, we inject collaborative commonsense into the slow model by transforming item-to-item knowledge into natural language explanations. A planner, trained through supervised warm-up followed by agentic reinforcement learning, dynamically decides which tool to invoke. Experiments on three datasets demonstrate that our method outperforms strong baselines, achieving consistent accuracy gains while reducing inference latency compared to uniform slow reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2605_11553
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TwiSTAR:Think Fast, Think Slow, Then Act,Generative Recommendation with Adaptive Reasoning
Cao, Shiteng
Jiang, Kaian
Gong, Yunlong
Li, Zhiheng
Information Retrieval
Generative recommendation with Semantic IDs (SIDs) has emerged as a promising paradigm, yet existing methods apply a fixed inference strategy, either fast direct generation or slow chain-of-thought reasoning, uniformly across all user histories. This approach creates a trade-off: fast recommendation model produces suboptimal accuracy on hard samples, while always invoking slow reasoning incurs prohibitive latency and wastes computation on easy cases. To address this, we propose Think Fast, Think Slow, Then Act, a framework that learns to adaptively allocate reasoning effort per user sequence. Our system equips an LLM with three complementary tools: a fast SID-based retriever, a lightweight candidate ranker, and a slow reasoning model that generates explicit rationales before recommending. Crucially, we inject collaborative commonsense into the slow model by transforming item-to-item knowledge into natural language explanations. A planner, trained through supervised warm-up followed by agentic reinforcement learning, dynamically decides which tool to invoke. Experiments on three datasets demonstrate that our method outperforms strong baselines, achieving consistent accuracy gains while reducing inference latency compared to uniform slow reasoning.
title TwiSTAR:Think Fast, Think Slow, Then Act,Generative Recommendation with Adaptive Reasoning
topic Information Retrieval
url https://arxiv.org/abs/2605.11553