SwapRec: Warming Up Cold Items Through Training-Time Swaps

_reachsumit · x · 2026-09-02

Addressing the fragility of sequential recommenders when swapping cold items with warm neighbors at inference time, this paper proposes SwapRec. It applies the same swap heuristics during training, making models robust to cold item interactions. Experiments in shopping, movie, and music domains show SwapRec significantly improves recommendation accuracy with cold interactions and increases the percentage of cold items recommended.

Original post →

More from Research

Research channel →