SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911605323726848 |
|---|---|
| author | Yu, Yang Kou, Lei Yi, Huaikuan Chen, Bin Cao, Yayu Shen, Lei Zhang, Chao Wang, Bing Zeng, Xiaoyi |
| author_facet | Yu, Yang Kou, Lei Yi, Huaikuan Chen, Bin Cao, Yayu Shen, Lei Zhang, Chao Wang, Bing Zeng, Xiaoyi |
| contents | With the rapid evolution of Large Language Models (LLMs), generative recommendation is gradually reshaping the paradigm of recommender systems. However, most existing methods remain confined to the interaction-driven next-item prediction paradigm, struggling to keep pace with the latest evolving trends or address the diverse recommendation tasks along with business-specific requirements in real-world scenarios. To this end, we present SIGMA, a Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender deployed at AliExpress. Specifically, we first ground item entities in a unified latent space capturing both general semantics and collaborative signals. Building upon this, we introduce a hybrid item tokenization method for both precise modeling and efficient generation. Moreover, we construct a large-scale multi-task supervised fine-tuning dataset empowering SIGMA to fulfill various recommendation demands via instruction-following. Finally, we design a three-step item generation procedure integrated with an adaptive probabilistic fusion mechanism to calibrate the output distributions based on task-specific requirements for recommendation accuracy and diversity. Extensive offline experiments and online A/B tests demonstrate the effectiveness of SIGMA across various real-world recommendation tasks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_22913 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress Yu, Yang Kou, Lei Yi, Huaikuan Chen, Bin Cao, Yayu Shen, Lei Zhang, Chao Wang, Bing Zeng, Xiaoyi Information Retrieval Machine Learning With the rapid evolution of Large Language Models (LLMs), generative recommendation is gradually reshaping the paradigm of recommender systems. However, most existing methods remain confined to the interaction-driven next-item prediction paradigm, struggling to keep pace with the latest evolving trends or address the diverse recommendation tasks along with business-specific requirements in real-world scenarios. To this end, we present SIGMA, a Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender deployed at AliExpress. Specifically, we first ground item entities in a unified latent space capturing both general semantics and collaborative signals. Building upon this, we introduce a hybrid item tokenization method for both precise modeling and efficient generation. Moreover, we construct a large-scale multi-task supervised fine-tuning dataset empowering SIGMA to fulfill various recommendation demands via instruction-following. Finally, we design a three-step item generation procedure integrated with an adaptive probabilistic fusion mechanism to calibrate the output distributions based on task-specific requirements for recommendation accuracy and diversity. Extensive offline experiments and online A/B tests demonstrate the effectiveness of SIGMA across various real-world recommendation tasks. |
| title | SIGMA: A Semantic-Grounded Instruction-Driven Generative Multi-Task Recommender at AliExpress |
| topic | Information Retrieval Machine Learning |
| url | https://arxiv.org/abs/2602.22913 |