Sequential Recommendation on Y-likes
1.53NDCG@10CAPE + Position-Aware Kernel Attention
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| CAPE + Position-Aware Kernel AttentionPositional Encoding Scheme=Combined CAPE and Kernel Attention2026.02 | 1.53 | 2.98 | 63.47 | |
| Position-Aware Kernel AttentionPositional Encoding Scheme=Proposed Kernel Attention2026.02 | 1.51 | 2.86 | 65.23 | |
| CAPEPositional Encoding Scheme=Contextual Positional Encoding2026.02 | 1.49 | 2.93 | 61.2 | |
| RoPEPositional Encoding Scheme=Rotary Positional Embeddings2026.02 | 1.48 | 2.85 | 48.82 | |
| NoPEPositional Encoding Scheme=No Positional Encoding2026.02 | 1.47 | 2.77 | 66.95 | |
| ClassicPositional Encoding Scheme=Additive Positional Embeddings2026.02 | 1.45 | 2.88 | 57.25 |