EchoRec: Multi-Item Prediction-Empowered Generative Recommendation via Cycle-Consistent Preference Alignment

2026-08-14

AI Summary

This study addresses the underutilization of future behaviors as dense supervision signals in multi-token prediction for generative recommendation. We propose EchoRec, which employs a cross-step cyclic consistency preference alignment mechanism to transform semantic echoes of future behaviors into effective supervision. By integrating stride-aware generation with a verifiable holistic alignment module, EchoRec theoretically precludes rank collapse-induced pseudo-alignment. Notably, auxiliary branches are removed during inference, ensuring zero additional computational overhead. Experimental results across three datasets demonstrate that EchoRec significantly enhances recommendation performance while inherently supporting efficient multi-item generation. This work effectively bridges the gap between leveraging future context and maintaining rigorous alignment in generative recommender systems.

Abstract

Generative recommendation autoregressively generates the semantic IDs of the target item, unifying preference modeling and index retrieval within the shared token space. Recent attempts have introduced Multi-Token Prediction (MTP) into this field, yet they primarily inherit its efficiency merit, leaving its potential as dense supervision unexplored. Unlocking this potential hinges on whether future behaviors qualify as informative supervision. Our analysis reveals that future behaviors carry a semantic echo of the current one far above that of random pairs, which nevertheless decays along horizons under intent transitions, making them informative yet order-dependent signals. Motivated by this, we propose EchoRec, which empowers MTP with cycle-consistent holistic preference alignment across multi-horizon for generative recommendation. It comprises two synergistic modules. Horizon-aware Preference Generation (HPG) sequentially chains lightweight auxiliary branches upon the base recommender, where each branch conditions on its predecessor to respect preference evolution. Verifiable Holistic-Preference Alignment (VHA) further consolidates them into the holistic preference and echoes it back through cycle-consistent projectors to suppress spurious alignment, with theoretical guarantees that exclude the rank-collapse form of spurious alignment under an invertible transport, enabling the holistic preference to be retained in the decoding representation. All auxiliary components serve as disposable scaffolding discarded at inference, introducing negligible online serving overhead. Extensive experiments on three datasets demonstrate the superiority of our EchoRec, together with its naturally acquired multi-item generation ability. Our code and datasets will be available upon acceptance.

Problem

• Generative Recommendation
• Multi-Token Prediction
• Preference Alignment
• Dense Supervision
• Future Behaviors

Innovation

• Multi-Token Prediction
• Cycle-Consistent Preference Alignment
• Generative Recommendation
• Holistic-Preference Alignment
• Horizon-aware Preference Generation