From Understanding to Action: Feedback-Grounded Policy Discovery for Generative Recommendation

2026-07-30

AI Summary

This work addresses the "understanding-action gap" in large language models (LLMs) when applied to recommender systems by proposing a feedback-driven agent framework. The approach first infers task-oriented user intent and then discovers effective recommendation strategies based on incremental utility and outcome feedback—rather than linguistic plausibility. It innovatively decouples the modeling of intent and policy knowledge, and compresses both into a lightweight semantic ID generator via dual-space relational distillation, enabling efficient LLM-free online inference. Evaluated on public benchmarks, the method significantly outperforms existing baselines, and large-scale online A/B tests demonstrate a 4.506% increase in revenue and a 4.621% improvement in ADVV (Average Daily Value per Visitor).

Abstract

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.

Problem

• Understanding-Action Gap
• Generative Recommendation
• Policy Discovery
• Intent Understanding
• Recommendation Policy

Innovation

• Understanding-Action Gap
• Feedback-Grounded Policy Discovery
• Intent-Policy Knowledge Distillation
• Semantic-ID Generator
• Dual-Space Relational Distillation