通过时间自回归对齐，解决生成式推荐早期搜索分支的误删问题。
TAAL: Mitigating Early Beam Pruning in Generative Recommendation via Temporal Autoregressive Alignment
- 训练时用前向KL对齐历史序列与早期前缀分布
- 推理时用PMI校准候选分数，减少高频前缀干扰
- 在亚马逊美妆等数据集上提升39.5%的NDCG@10
生成式推荐将物品编码为层次语义标识符（SIDs），并通过自回归解码预测下一个物品。然而，标准的下一步词预测未显式建模交互序列中的多模态转移，导致真实SID在早期束搜索分支中极易被不可逆地剪枝。在三个公开基准上，我们发现91.9%–96.6%的检索失败发生在前两步解码中。为此，提出时间自回归对齐（TAAL）。训练时，TAAL从历史转移构建联合$(c_1,c_2)$软目标，并通过前向KL目标对齐早期前缀分布；推理时，使用点互信息（PMI）校准候选分数，降低全局高频前缀的影响。在Amazon Beauty、Instruments和Yelp上，TAAL相较标准基线在NDCG@10上分别提升39.5%、6.7%和28.6%，同时完整SID存活率提高3.9%–16.6%。束宽分析显示，随着束宽减小，相对存活率提升更大，在B=5时达到39.4%。
原文摘要 · Abstract (English)
Generative recommendation encodes items as hierarchical semantic identifiers (SIDs) and retrieves the next item through autoregressive decoding. Standard next-token prediction, however, does not explicitly cover the multimodal transitions present in interaction sequences, leaving the ground-truth SID vulnerable to irreversible pruning at early beam-search branches. Across three public benchmarks, we find that 91.9\%--96.6\% of retrieval failures occur within the first two decoding steps. We therefore propose Temporal Autoregressive Alignment (TAAL). During training, TAAL constructs a joint $(c_1,c_2)$ soft target from historical transitions and aligns the early-prefix distribution with a forward KL objective. During inference, it calibrates candidate scores with pointwise mutual information (PMI) to reduce the influence of globally frequent prefixes. On Amazon Beauty, Instruments, and Yelp, TAAL improves NDCG@10 over the standard baseline by 39.5\%, 6.7\%, and 28.6\%, respectively, while increasing full-SID survival by 3.9\%--16.6\%. Beam-width analysis further shows that the relative survival gain grows as the beam narrows, reaching 39.4\% at $B=5$.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。