LoopRec: Item-Level Adaptive Recursion for Sequential Recommendation

Forum topic · 小凯 · 2026-10-23

Summary

Researchers propose LoopRec, a method that replaces fixed-depth computation in sequential recommenders with item-level adaptive recursion. The authors present two diagnostic findings: computational demand across items in a user's interaction history is highly non-uniform—most items converge after shallow processing while a critical minority requires full depth—and fixed uniform allocation is suboptimal, as oracle-based non-uniform depth assignment consistently outperforms fixed allocation. LoopRec recursively applies a single weight-sharing block and inserts a lightweight gating mechanism at each recursion boundary that makes a binary exit-or-continue decision per item. Mature representations exit early; harder items continue deeper recursion. This converts fixed global depth into a learnable per-item property without auxiliary depth labels and adapts easily to multiple backbone architectures. Experiments on public benchmarks across four backbone families show LoopRec consistently outperforms matched fixed-depth counterparts. Learned allocation patterns are interpretable: low-frequency items and more recent interactions naturally receive deeper computation. Paper: arXiv 2610.30128, posted October 2026.

Paper Overview

Field: NLP / Sequential Recommendation Authors: Zhanshuo Cao, Jiakai Tang, Yuqi Zhou, Sunhao Dai, Jun Xu Posted: 2026-10-21 arXiv: 2610.30128

Abstract (Translated)

Existing sequential recommenders usually allocate a fixed computational budget to every item in a user interaction history, regardless of the item's contribution to the final prediction. The authors challenge this de facto design with two diagnostic findings:

1. Non-uniform computational need: items within the same sequence vary drastically in how much computation they require. Most converge after shallow processing, while a critical minority only stabilizes at full depth. 2. Suboptimality of fixed allocation: given oracle access to the target item, a non-uniform depth assignment consistently outranks the fixed one, indicating that existing uniform computation both wastes capacity on already-converged items and sacrifices overall recommendation quality.

These observations motivate a shift from fixed-depth computation to item-level adaptive inference. The authors propose LoopRec, which recursively applies a single weight-sharing block and introduces a lightweight gating mechanism at each recursion boundary to make an independent binary decision—exit or continue—for each item. Items whose representations have fully matured exit early, while items requiring deeper reasoning proceed through subsequent recursions.

This design turns fixed global depth into a learnable per-item attribute, requires no auxiliary depth labels, and can be easily adapted to multiple backbone architectures.

Results

Extensive experiments on public benchmarks across four backbone architecture families show that LoopRec consistently outperforms matched fixed-depth counterparts.
Further analysis reveals interpretable allocation patterns: low-frequency items and more recent interactions naturally receive deeper computation.