Towards End-to-End Alignment of User Satisfaction via Questionnaire in Video Recommendation

2026-01-28

AI Summary

This work addresses the challenge of incorporating sparse but reliable explicit user satisfaction signals—collected via questionnaires—into large-scale short-video recommendation systems, where conventional approaches rely on noisy implicit signals such as clicks and watch time. To overcome the sparsity and integration difficulties of explicit feedback, the authors propose EASQ, a novel framework that enables end-to-end incorporation of explicit satisfaction signals in online recommendation. EASQ employs multi-task learning with parameter isolation via lightweight LoRA modules to construct dedicated pathways for processing sparse questionnaire data, and introduces a DPO-based optimization objective tailored for online learning to prevent the explicit signals from being overwhelmed by dense behavioral data. Both offline evaluations and online A/B tests demonstrate that EASQ significantly improves user satisfaction across multiple scenarios and has been successfully deployed in production, yielding substantial business gains.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Search and Optimization: Learning to Search
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Short-video recommender systems typically optimize ranking models using dense user behavioral signals, such as clicks and watch time. However, these signals are only indirect proxies of user satisfaction and often suffer from noise and bias. Recently, explicit satisfaction feedback collected through questionnaires has emerged as a high-quality direct alignment supervision, but is extremely sparse and easily overwhelmed by abundant behavioral data, making it difficult to incorporate into online recommendation models. To address these challenges, we propose a novel framework which is towards End-to-End Alignment of user Satisfaction via Questionaire, named EASQ, to enable real-time alignment of ranking models with true user satisfaction. Specifically, we first construct an independent parameter pathway for sparse questionnaire signals by combining a multi-task architecture and a lightweight LoRA module. The multi-task design separates sparse satisfaction supervision from dense behavioral signals, preventing the former from being overwhelmed. The LoRA module pre-inject these preferences in a parameter-isolated manner, ensuring stability in the backbone while optimizing user satisfaction. Furthermore, we employ a DPO-based optimization objective tailored for online learning, which aligns the main model outputs with sparse satisfaction signals in real time. This design enables end-to-end online learning, allowing the model to continuously adapt to new questionnaire feedback while maintaining the stability and effectiveness of the backbone. Extensive offline experiments and large-scale online A/B tests demonstrate that EASQ consistently improves user satisfaction metrics across multiple scenarios. EASQ has been successfully deployed in a production short-video recommendation system, delivering significant and stable business gains.

Problem

• user satisfaction
• questionnaire
• video recommendation
• sparse feedback
• behavioral signals

Innovation

• questionnaire-based feedback
• LoRA
• multi-task learning
• DPO
• end-to-end alignment