MARS通过多尺度时间信号聚合，提升稀疏与密集数据下的推荐精度。
MARS: Multi-rate Aggregation of Recency Signals for Sequential Recommendation across Sparse and Dense Regimes
- 基于真实时间戳生成多尺度历史行为摘要，自适应融合
- 在稀疏数据上相对最强基线提升36.2%的HR@10，密集数据效率更高
- 适配不同数据密度，可无缝替换Transformer或Mamba编码器
序列推荐模型通常通过位置自注意力（如Transformer）或单一隐式衰减策略（如状态空间模型）加权历史行为，但未显式建模用户行为的时间多尺度结构。本文提出MARS，一种编码器无关的聚合算子，利用真实时间戳生成K个强调不同近期性的摘要，并通过上下文自适应门控融合。MARS参数增加不超过6%，计算复杂度为O(LdK)。它根据训练集平均序列长度自动选择：稀疏数据用MARS-T（Transformer），密集数据用MARS-M（Mamba）。在五个公开基准上，对比十种基于Transformer和Mamba的基线，在统一RecBole协议下，MARS在所有任务中均取得最佳HR@10，稀疏数据上相较最强内容仅基线平均提升19.7%（游戏数据达+36.2%）；密集数据上相较SIGMA提升+3.2% HR@10 / +0.9% NDCG，且仅需42%的MFLOPs，占据跨数据密度的准确率-效率前沿。纯骨干消融实验表明，MARS贡献为稀疏数据上+4%至+19%的HR@10，验证了双实例化设计的有效性。代码见补充材料。
原文摘要 · Abstract (English)
Sequential recommenders weight historical interactions either through positional self-attention as in Transformers or through a single implicit decay schedule as in State-Space Models. Neither makes the multi-scale temporal structure of real user behaviour explicit. We propose MARS, an encoder-agnostic aggregation operator that consumes real timestamps and produces K summaries emphasising distinct recency scales, fused by a context-adaptive gate. MARS adds at most 6% parameters and runs in $\mathcal{O}(LdK)$ time. MARS adapts to data density by automatically selecting between two encoder instantiations: MARS-T (Transformer) for sparse data and MARS-M (Mamba) for dense data, based on the average sequence length of the training set. On five public benchmarks against ten Transformer- and Mamba-based baselines under a unified RecBole protocol, MARS attains the best HR@10 on every benchmark, with mean relative gain +19.7% over the strongest content-only Transformer baseline on sparse data (reaching +36.2% on Games) and +3.2% HR@10 / +0.9% NDCG over SIGMA on dense ML-1M at 42% fewer MFLOPs, occupying the accuracy-efficiency Pareto frontier across the data-density spectrum. A backbone-only ablation isolates the marginal contribution of MARS at +4% to +19% HR@10 on sparse data and motivates the dual-instantiation design. The code is included in the supplementary material.
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。