发表机构
National University of Singapore; Beijing University of Posts and Telecommunications; Tencent; Shandong University（新加坡国立大学; 北京邮电大学; 腾讯; 山东大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
EchoRec通过跨多时间步长的循环一致整体偏好对齐，结合感知时间步长的偏好生成与可验证的整体偏好对齐模块，在三个数据集上实现了更优的生成式推荐性能，且具备多物品生成能力。
AI 中文摘要
生成式推荐通过自回归方式生成目标物品的语义ID，在共享token空间内统一了偏好建模与索引检索。近期研究将多token预测（Multi-Token Prediction, MTP）引入该领域，但主要利用其效率优势，未探索其作为密集监督信号的潜力。挖掘该潜力的关键在于未来行为是否可作为有效监督信号。我们的分析显示，未来行为携带当前行为的语义回声，远高于随机配对的情况，不过在意图转变时会随时间步长衰减，因此是有效但依赖顺序的信号。受此启发，我们提出EchoRec，通过跨多时间步长的循环一致整体偏好对齐，为生成式推荐赋能多token预测（MTP）。该模型包含两个协同模块：感知时间步长的偏好生成（Horizon-aware Preference Generation, HPG）在基础推荐器上依次串联轻量辅助分支，每个分支依赖前一分支以尊重偏好演化；可验证的整体偏好对齐（Verifiable Holistic-Preference Alignment, VHA）进一步将这些分支整合为整体偏好，并通过循环一致投影器将其回声反馈以抑制虚假对齐，理论保证在可逆变换下排除秩崩溃形式的虚假对齐，使整体偏好保留在解码表示中。所有辅助组件在推理时作为一次性支架被丢弃，仅引入可忽略的在线服务开销。在三个数据集上的大量实验证明了EchoRec的优越性，以及其自然获得的多物品生成能力。我们的代码和数据集将在录用后发布。
英文摘要
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.
Comments10 pages, 9 figures, Under Review
