Search papers, labs, and topics across Lattice.
This paper addresses the Understanding-Action Gap in generative recommendation systems by introducing a feedback-driven agent framework that differentiates between intent knowledge and policy knowledge. By leveraging outcome-derived feedback to refine recommendation policies, the framework enhances the effectiveness of item suggestions beyond what traditional LLMs can achieve. Experimental results demonstrate significant improvements in revenue and user engagement metrics, validating the approach's practical utility in real-world applications.
Bridging the Understanding-Action Gap leads to a 4.5% revenue increase by refining recommendation policies through direct user feedback rather than linguistic reasoning alone.
Semantic-ID-based generative recommenders enable efficient next-item generation, but their item-level supervision mainly captures behavioral co-occurrence and local transitions. Large language models (LLMs) can complement these models by reasoning over heterogeneous interaction histories to understand the user's current demand. However, LLMs are not inherently trained with recommendation-specific outcome feedback, and linguistically plausible reasoning therefore does not necessarily lead to effective recommendation decisions. We term this mismatch the Understanding-Action Gap. Accordingly, we distinguish intent knowledge, which captures the user's current demand, from policy knowledge, which specifies the recommendation direction and rejection boundary under that demand. To bridge this gap, we propose a feedback-driven agent framework that first induces task-oriented intent and then discovers recommendation policies according to their incremental utility over an intent-only baseline. Candidate policies are evaluated and refined using outcome-derived feedback rather than linguistic plausibility. We further transfer the resulting intent and policy knowledge into two latent tokens of a lightweight Semantic-ID generator through dual-space relational distillation, enabling LLM-free online inference. Experiments on public benchmarks show consistent improvements over baselines, while large-scale online A/B tests achieve gains of 4.506% in Revenue and 4.621% in ADVV.
