Dual-Enhancement Product Bundling: Bridging Interactive Graph and Large Language Model

Zhe Huang; Peng Wang; Yan Zheng; Sen Song; Longjun Cai

cs.CLApr 15, 2026

Dual-Enhancement Product Bundling: Bridging Interactive Graph and Large Language Model

Authors: Zhe Huang, Peng Wang, Yan Zheng, Sen Song, Longjun Cai

Abstract

Product bundling boosts e-commerce revenue by recommending complementary item combinations. However, existing methods face two critical challenges: (1) collaborative filtering approaches struggle with cold-start items owing to dependency on historical interactions, and (2) LLMs lack inherent capability to model interactive graph directly. To bridge this gap, we propose a dual-enhancement method that integrates interactive graph learning and LLM-based semantic understanding for product bundling. Our method introduces a graph-to-text paradigm, which leverages a Dynamic Concept Binding Mechanism (DCBM) to translate graph structures into natural language prompts. The DCBM plays a critical role in aligning domain-specific entities with LLM tokenization, enabling effective comprehension of combinatorial constraints. Experiments on three benchmarks (POG, POG_dense, Steam) demonstrate 6.3%-26.5% improvements over state-of-the-art baselines.

Figures & tables

Figure 1: An example of product bundling.

Figure 2: Different methods illustration on product bundling tasks.

Figure 3: The dual-enhancement framework for product bundling.

Figure 4: Illustration of dynamic concept binding mechanism.

Datasets	#U	#I	#B
POG	17,449	48,676	20,000
POG_dense	2,311,431	31,217	29,686
steam	54,985	2,798	615

Table 1: The statistics of the three datasets on two different domains.

Dataset	POG		POG_dense		steam
Model	HitRate@1	ValidRatio	HitRate@1	ValidRatio	HitRate@1	ValidRatio
ICL †	0.10	1.00	0.08	1.00	0.12	0.13
ICL ‡	0.10	1.00	0.10	1.00	0.57	0.83
ICL ⋈	0.23	0.79	0.20	0.78	0.76	0.98
MultiVAE	0.28	1.00	0.45	1.00	-	-
BiLSTM	0.29	1.00	0.43	1.00	-	-

Table 2: The results of DPB-LLM compared with baseline methods.

Dataset	POG		POG_dense		steam
Model	HitRate@1	ValidRatio	HitRate@1	ValidRatio	HitRate@1	ValidRatio
PB-LLM w/o GCN †	0.44	1.00	0.65	1.00	0.75	1.00
PB-LLM †	0.46	1.00	0.67	1.00	0.76	1.00
DPB-LLM w/o GCN †	0.48	1.00	0.70	1.00	0.80	1.00
DPB-LLM-JT †	0.43	1.00	0.65	1.00	0.77	1.00
DPB-LLM †	0.49	1.00	0.71	1.00	0.81	1.00

Table 3: Ablation studies related to DPB-LLM.

Figure 5: Comparison of different LLM parameter scales.

Figure 6: Comparison of different numbers of candidate items.

Large Language Model-Based Recommendation

Large language model-based recommendation uses language models to personalize ranking, retrieval, and conversational suggestions from text and behavior.

See all 210 papers

Hypergraph-Enhanced Dual Convolutional Network for Bundle Recommendation
Dec 18, 2023Yang Li, Kangbo Liu, Yaoxin Wu +2Recommender Systems Hypergraph Neural Networks
Northwestern Polytechnical University, China · Nanyang Technological University, Singapore
Enhancing High-order Interaction Awareness in LLM-based Recommender Model
Sep 30, 2024Xinfeng Wang, Jin Cui, Fumiyo Fukumoto +1Sequential Recommendation Feature Interaction Modeling
Graduate School of Engineering University of Yamanashi, Kofu, Japan · Interdisciplinary Graduate School University of Yamanashi, Kofu, Japan
DCGL: Dual-Channel Graph Learning with Large Language Models for Knowledge-Aware Recommendation
May 8, 2026Xinchi Zou, Tongzhenzhi Su, Jianjun Li +4Recommender Systems Large Language Model-Based Recommendation
School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China · School of Software Engineering, Huazhong University of Science and Technology, Wuhan, China · Laboratory for Artificial Intelligence and New Forms of Education, Central China Normal University, Wuhan, China

Dec 18, 2023J/K move · Enter open · S save

Hypergraph-Enhanced Dual Convolutional Network for Bundle Recommendation

Yang Li, Kangbo Liu, Yaoxin Wu, Zhaoxuan Wang+1

Northwestern Polytechnical University, China · Nanyang Technological University, Singapore

Recommender Systems Hypergraph Neural Networks

Open paper AskPDF arXiv

Bundle recommendation ranks sets of related items rather than isolated items. Its central challenge is to connect user preferences, item interactions, and bundle composition without losing the signals needed to rank bundles. We propose Hypergraph-Enhanced Dual Convolutional Neural Network (HED), which constructs a complete hypergraph containing user--bundle, user--item, and bundle--item interactions together with intra-user and intra-bundle relations. HED couples complete-hypergraph propagation with a user--bundle branch, allowing item-aware higher-order context to inform ranking while preserving recommendation-specific signals. On NetEase, HED-128 improves over the strongest baseline by 5.04--6.97% across the six reported metrics; on Youshu, HED-64 improves by 1.87--4.56%. Ablation results support the contributions of both the user--bundle branch and intra-type relations, and sensitivity analyses identify stable operating ranges for the main hyperparameters. We further quantify the computational trade-off of the complete hypergraph, including its memory cost. The evidence supports HED on the two evaluated bundle-recommendation datasets while making its resource limitations explicit. Code and datasets will be made available upon publication.

Abstract

Large Language Model-Based Recommendation

Explore similar work

Hypergraph-Enhanced Dual Convolutional Network for Bundle Recommendation

Enhancing High-order Interaction Awareness in LLM-based Recommender Model

DCGL: Dual-Channel Graph Learning with Large Language Models for Knowledge-Aware Recommendation