Real-Time Trend Prediction via Continually-Aligned LLM Query Generation

2026-01-24

AI Summary

This work addresses the cold-start challenge in detecting emerging and long-tail trends under low-traffic search scenarios, where query sparsity hinders timely trend identification. The authors propose a continual learning framework based on large language models (CL-LLM), employing a Mix-Policy Direct Preference Optimization (DPO) method to mitigate catastrophic forgetting while preserving reasoning capabilities. To bypass reliance on real user queries, they synthesize high-quality search queries from news content and devise a trend scoring mechanism that integrates user interaction intensity and creator authority for early trend detection. Deployed in Facebook and Meta AI products, the system achieves a 91.4% improvement in long-tail trend detection precision@500, a 19% gain in query generation accuracy, and sustained performance stability over multiple weeks of online training.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Life-Long and Continual Learning
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Trending news detection in low-traffic search environments faces a fundamental cold-start problem, where a lack of query volume prevents systems from identifying emerging or long-tail trends. Existing methods relying on keyword frequency or query spikes are inherently slow and ineffective in these sparse settings, lagging behind real-world shifts in attention. We introduce RTTP, a novel Real-Time Trending Prediction framework that generates search queries directly from news content instead of waiting for users to issue them. RTTP leverages a continual learning LLM (CL-LLM) that converts posts into search-style queries and scores them using engagement strength + creator authority, enabling early trend surfacing before search volume forms. To ensure adaptation without degrading reasoning, we propose Mix-Policy DPO, a new preference-based continual learning approach that combines on-policy stability with off-policy novelty to mitigate catastrophic forgetting during model upgrades. Deployed at production scale on Facebook and Meta AI products, RTTP delivers +91.4% improvement in tail-trend detection precision@500 and +19% query generation accuracy over industry baselines, while sustaining stable performance after multi-week online training. This work demonstrates that LLM-generated synthetic search signals, when aligned and continually updated, unlock timely trend understanding in low-traffic search environments.

Problem

• cold-start problem
• trending news detection
• low-traffic search
• long-tail trends
• query volume

Innovation

• Continual Learning
• LLM Query Generation
• Trend Prediction
• Mix-Policy DPO
• Cold-Start Problem