PIANO: Personalized Reranking via Information Aggregation Node for Music Search Optimization

2026-06-15

AI Summary

This work addresses the challenge of music search re-ranking, which requires aligning both the user’s current query and long-term preferences while jointly optimizing click-through rate (CTR) and conversion rate (CVR) at the list level. Existing approaches struggle to effectively leverage historical queries for intent matching and are limited to item-level multi-objective optimization. To overcome these limitations, we propose the PIANO framework, which first aligns historical and current search intents via a cross-attention-based Query-Driven Interest Refiner, and then introduces a learnable, [CLS]-like Information Aggregation Node to enable end-to-end list-wise joint modeling of CTR and CVR. PIANO is the first method to utilize historical search queries for intent alignment and to achieve list-level multi-objective optimization in an end-to-end manner. Experiments show that PIANO significantly outperforms strong baselines on both public and industrial datasets, with online A/B tests on NetEase Cloud Music demonstrating a 0.62% CTR gain and a 4.45% CVR improvement.

Abstract

Unlike short-video content, music tracks have long lifecycles and lasting value. Effective music search re-ranking must therefore align the user's current query with long-term preferences while jointly optimizing Click-Through Rate (CTR) and Conversion Rate (CVR). However, existing methods suffer from two limitations: (1) sequential methods rely on item-interaction history and therefore cannot use historical search queries to tell which past preferences match the user's current search intent; (2) most listwise models optimize a single objective (e.g., CTR only), and conventional multi-objective methods balance click and conversion at the item level, ignoring how these trade-offs play out across the whole ranked list. To address these limitations, we propose PIANO, a personalized listwise re-ranking framework with two key components: (i) the Query-Driven Interest Refiner (QDIR) uses cross-attention over historical queries to align past intents with the current one; (ii) the Information Aggregation Node (IAN), a learnable [CLS]-style token, aggregates the candidate list and predicts CTR/CVR at the list level. Extensive experiments on public and industrial datasets show consistent gains over strong baselines. In online A/B tests on NetEase Cloud Music, a leading music streaming platform, PIANO achieves statistically significant improvements in CTR (+0.62%) and CVR (+4.45%).

Problem

• music search re-ranking
• personalization
• multi-objective optimization
• CTR
• CVR

Innovation

• personalized re-ranking
• information aggregation node
• query-driven interest refinement
• listwise multi-objective optimization
• music search