EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

BiFlowRec: Bidirectional Rectified Flow for LLM-Conditional Generative Recommendation

Shaohua Li, Qi Huang

IEEE Access · 2026 · 14 · 74156-74169

DOI: 10.1109/access.2026.3690573 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Shaohua Li und 1 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Major Science and Technology Project of Xinjiang Uygur Autonomous Region “Efficient and Low-Cost Safe Utilization Technology of Brackish Water”.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Generative recommendation models user preferences over the entire item set. With the rapid advances of large language models (LLMs), an emerging challenge is how to let LLM semantics directly shape this generative process rather than merely serving as side features. Existing approaches to LLM-conditioned generative recommendation typically perform distribution alignment in a VAE latent space, which introduces an encode–align–decode pipeline and can incur information loss and additional computational overhead. We propose $\textsf {BiFlowRec}$ , a Rectified Flow-based symmetric bidirectional conditional flow matching model that directly constructs LLM-conditioned vector fields in the interaction space for both user $\rightarrow $ item and item $\rightarrow $ user directions. $\textsf {BiFlowRec}$ employs semantics-aware confidence weighting to suppress noisy interactions and leverages the straight-line property of Rectified Flow to support single-step generation. Experiments on three public datasets (Amazon-Book, Yelp, and Steam) show that $\textsf {BiFlowRec}$ , as an end-to-end generative recommender, consistently outperforms a wide range of generative and LLM-enhanced baselines. In particular, $\textsf {BiFlowRec}$ improves Recall@10 by 34.0%, 6.8%, and 2.3% on the three datasets, respectively, over the strongest generative baselines. Ablation studies further validate the effectiveness of the symmetric bidirectional architecture, the semantics-aware sample reweighting strategy, and the overall training design.

Förderung

  • Major Science and Technology Project of Xinjiang Uygur Autonomous Region “Efficient and Low-Cost Safe Utilization Technology of Brackish Water”

Zitieren

APA

Li, S., & Huang, Q. (2026). BiFlowRec: Bidirectional Rectified Flow for LLM-Conditional Generative Recommendation. IEEE Access, 14, 74156-74169. https://doi.org/10.1109/access.2026.3690573

Vancouver

Li S, Huang Q. BiFlowRec: Bidirectional Rectified Flow for LLM-Conditional Generative Recommendation. IEEE Access. 2026;14:74156-74169. doi:10.1109/access.2026.3690573

BibTeX
@article{Li2026,
  author = {Li, Shaohua and Huang, Qi},
  title = {BiFlowRec: Bidirectional Rectified Flow for LLM-Conditional Generative Recommendation},
  journal = {IEEE Access},
  year = {2026},
  volume = {14},
  pages = {74156--74169},
  doi = {10.1109/access.2026.3690573}
}
Kostenlos starten →