When Top-ranked Recommendations Fail: Modeling Multi-Granular Negative Feedback for Explainable and Robust Video Recommendation

2025-11-23

AI Summary

Existing video recommendation systems rely on ID embeddings and collaborative filtering, limiting their ability to model video semantics and disentangle user biases—such as accidental clicks or skimming—leading to frequent, unattributable negative feedback. Method: We introduce the first real-world user viewing sequence dataset annotated with fine-grained “dislike reasons,” and propose the Agentic Explainable Negative Feedback (AENF) framework. AENF employs three specialized agents—Profile, Video, and Reason—to jointly model user interests and aversion mechanisms, and integrates S-GRPO, a reinforcement fine-tuning algorithm, to enable multi-granularity negative feedback attribution and explainable recommendations. The approach synergizes multimodal video understanding, behavioral sequence modeling, psychological profiling, and agent-based reasoning. Contribution/Results: AENF achieves an 8.6% higher negative feedback attribution accuracy than GPT-4o. Online deployment yields a 6.2% increase in average watch time, a 9.4% reduction in skip rate, and significant improvements in user satisfaction.

Technology Category

• Computer Vision: Video Understanding & Activity Analysis
• Multiagent Systems: Adversarial Agents
• Knowledge Representation and Reasoning: Preferences

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Existing video recommendation systems, relying mainly on ID-based embedding mapping and collaborative filtering, often fail to capture in-depth video content semantics. Moreover, most struggle to address biased user behaviors (e.g., accidental clicks, fast skips), leading to inaccurate interest modeling and frequent negative feedback in top recommendations with unclear causes. To tackle this issue, we collect real-world user video-watching sequences, annotate the reasons for users' dislikes, and construct a benchmark dataset for personalized explanations. We then introduce the Agentic Explainable Negative Feedback (ENF) framework, which integrates three core components: (1) the Profile Agent, extracting behavioral cues from users' historical data to derive psychological and personality profiles; (2) the Video Agent, performing comprehensive multimodal video analysis; and (3) the Reason Agent, synthesizing information from the other two agents to predict user engagement and generate explanations. Additionally, we propose the S-GRPO algorithm, enabling the model to progressively address complex tasks during reinforcement fine-tuning. Experimental results on the collected dataset show that our method significantly outperforms state-of-the-art baselines in negative feedback prediction and reason explanation. Notably, it achieves an 8.6% improvement over GPT-4o in reason classification. Deployment on the business platform further validates its benefits: increasing average user watch time by 6.2%, reducing the fast-skip rate by 9.4%, and significantly enhancing user satisfaction.

Problem

• Video recommendations fail to capture content semantics and user intent
• Systems struggle with biased behaviors like accidental clicks and skips
• Current methods provide unclear causes for negative feedback in recommendations

Innovation

• Agentic framework integrates profile, video, and reason agents
• S-GRPO algorithm enables progressive reinforcement fine-tuning
• Multimodal analysis synthesizes behavioral cues and video content