Unified ranking model improves multi-task recommendations
Task-Blind No MORE: Multi-Task Information Flow in Unified Ranking Backbones
Information Retrieval
Summary
Many recommendation systems suggest items by considering different goals, like clicks or time spent, but often treat these goals separately. The authors introduce MORE, a model that mixes signals for all tasks throughout the whole system, not just at the end. It uses special 'Anchor Tokens' to share what’s common and keep what’s different between tasks, improving overall recommendations. Tests on big data and a popular app show this approach helps people interact more and makes the system run faster.
What this means in practice
- •For recommender system engineers: Build recommendation models that improve multiple user engagement metrics by integrating task signals throughout model layers instead of only after backbone processing.
- •For social media platform teams: Enhance user interaction and session duration on apps by deploying multi-task ranking models that efficiently share computations and run faster.$Commercial implications: Enables commercially competitive recommendation services increasing user engagement in social apps through better multi-task modeling and reduced latency.
Authors
Yuchen Wang, Feng Niu, Qing Tan, Junting Lu, Baoxin Wu, Jun Gao
Abstract
Industrial ranking models for recommendation have scaled feature interaction and sequence modeling separately; recent architectures such as HyFormer and MixFormer unify both in a stackable backbone. Real-world recommender systems, however, nearly always require multi-task learning, yet existing unified architectures confine multi-task modeling to shallow post-backbone towers, leaving the backbone without task-aware information flow. We propose MORE (Multi-task cO-evolving Ranking modEl), which embeds multi-task information flow inside the backbone, enabling task-specific signals to co-evolve with sequence and feature representations at every layer rather than in a post-hoc fusion. It introduces Anchor Tokens that persist across backbone layers: Shared Anchors encode cross-task commonalities, while Private Anchors capture task-specific priors. In each block, Anchor Tokens (1) read task-conditioned signals from behavior sequences, (2) mix with non-sequential features under a task-boundary mask, and (3) refine per-task representations through independent branches; as blocks stack, each task obtains a differentiated representation refined through all backbone layers. Experiments on large-scale industrial datasets show that MORE consistently outperforms baselines across all tasks under comparable parameter and FLOPs budgets, and scales well with model size. Online A/B tests on Momo, a leading Chinese social discovery platform with tens of millions of monthly active users, yield 3% improvement in usage duration, 3.6% in interaction rate, and 2% in deep-chat rate. MORE is deployed in production with request-level shared computation reducing scoring latency by about 30%.