LoopRec assigns per-item recursion depth in sequential recommenders
Original title: Think Harder Where It Matters: Item-Level Adaptive Recursion for Sequential Recommendation
A shared block and per-item exit gate beat matched fixed-depth models; rarer and more recent items get deeper compute.
LoopRec replaces a fixed computation budget in sequential recommendation with item-level adaptive recursion. A single weight-shared block is applied repeatedly, and a lightweight gate decides independently whether each item exits or continues. The paper says the method needs no auxiliary depth labels and works across four backbone families on public benchmarks, consistently beating matched fixed-depth models. Learned allocation gives more depth to lower-frequency items and more recent interactions, matching the authors' oracle analysis. Code is released at the stated GitHub repository.
- Each item gets an independent binary exit-or-continue decision.
- No auxiliary depth labels are used.
- Tests cover four backbone families on public benchmarks.
- Lower-frequency and more recent items receive deeper computation.
- Code is at github.com/Elpmis117/LoopRec.
- AI models
- LoopRec
From the original article
1 min read at huggingface.co
Existing sequential recommenders usually allocate a fixed computational budget to every item in a user's interaction history, regardless of the item's contribution to the final prediction. We 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.
By Zhanshuo Cao, Jiakai Tang, Yuqi Zhou · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.