New method boosts user modeling by using negative behavior signals

Researchers have developed a new method for sequential user modeling in click-through rate prediction that incorporates implicit negative behaviors, such as skips and scroll-pasts, alongside positive interactions. This approach, termed Target-Aware Polarity Fusion (TAPF), uses a lightweight gating mechanism to differentiate behavioral evidence and has shown consistent improvements of up to 9.6% in relative AUC across various model architectures. The study highlights that the primary contribution lies in the mixed-polarity data paradigm itself, which significantly outperforms models relying solely on positive signals. AI

IMPACT Enhances user modeling accuracy by incorporating previously underutilized negative behavioral signals.

RANK_REASON The cluster contains a research paper published on arXiv detailing a new method for sequential user modeling. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.IR (Information Retrieval) →

• alphaXiv
• arXiv
• CatalyzeX
• Click Through Rate
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• information retrieval
• ScienceCast
• Target-Aware Polarity Fusion

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →