Diffusion model boosts diverse and faster item bundle recommendations
DPE-Rec: Diversity-Preserving and Efficient Diffusion Language Model Recommendation for Bundle Construction
Information Retrieval
Summary
When online stores suggest groups of products that go well together, it's important these suggestions are both diverse and made quickly. The authors found that special language models called diffusion models can generate such product bundles more naturally than traditional methods. They introduced a new approach that keeps the variety of suggestions high while making the recommendation process faster. Their tests show big improvements in recommendation quality and speed compared to existing methods.
What this means in practice
- •For e-commerce platform engineers: Generate diverse and faster product bundle recommendations from large catalogs to enhance customer experience.
- •For online retail analytics teams: Improve the quality and efficiency of bundle recommendation algorithms using diffusion-based language models.
Authors
Pingchuan Ma, Bohan Jiang, Chengshuai Zhao, Dawei Li, Zhen Tan, Huan Liu
Abstract
Generative bundle construction completes a partially observed bundle by generating its missing items from a large product catalog. Compared to autoregressive LLMs, diffusion LLM emerges as a natural fit for this unordered generation task. Their bidirectional attention provides an inductive bias better aligned with unordered bundle construction. Iterative denoising can reveal multiple tokens per step, potentially reducing inference latency. However, directly applying DLMs to this task presents two key challenges: 1 Despite its flexible reveal order, confidence-guided decoding tends to postpone semantic-fork decisions until refinement tokens constrain their choices, reducing bundle diversity. 2 Revealing more tokens per step can compromise recommendation quality, stifling the potential for faster inference. We introduce DPE-Rec, a diffusion framework for diversity preservation and efficient recommendation. DPE-Rec preserves diversity by introducing semantic-fork-aware GRPO, which prioritizes semantic-fork decisions via autoregressive rollouts. It improves efficiency through pseudo-trajectory distillation, which concentrates learning from the vast random corruption space onto the compact trajectory-induced subspace visited by high-quality decoding, unlocking reliable multi-token decoding. Together, these components enable diverse and accurate bundle construction with substantially fewer denoise steps. Experiments on multiple bundle construction benchmarks show that DPE-Rec improves recommendation performance by 63% on average, and its fast variant achieves 25.6$times$ faster inference on average than existing generative recommendation arts, with minimal loss in recommendation quality.