Research area

User reactivation, generative recommendation, Semantic-ID interface diagnostics, dynamic retrieval, graph embedding, counterfactual watch-time prediction, and efficient CTR modeling form the main technical line.

Publication record

Authors

Fuyuan Liu, Tiandeng Wu, Yaqun Fang, Wei Zhou, Zehao Zhou, Wenping Chen, Qishun Mei, Jiaxin Zhou, Heng Chang, Yi Cao, Jiandong Ding

Venue

Proceedings of the ACM International Conference on Information and Knowledge Management (CIKM '26)

Year

2026

Area

Recommender Systems

Problem

Multi-task recommendation can suffer when learning signals from one task erode or distort the signal needed by another task.

Approach

The paper studies personalized task dependency graphs as a way to model task relationships at a finer granularity than a single global dependency structure.

Public evidence

The public paper reports AUC gains of up to 1.45% on sparse conversion tasks, together with online improvements of 1.2% in conversion rate and 1.9% in effective cost per mille relative to the production baseline.

Takeaway

The work belongs to the recommendation line on reliable multi-objective modeling under changing user, item, and task signals.