# Bursting Filter Bubble: Enhancing Serendipity Recommendations with Aligned Large Language Models

arXiv: https://arxiv.org/abs/2502.13539 · HTML: https://mlsanity.com/paper/2502.13539/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 19 Feb 2025.

## Built on by (most cited first, 6 of 6)

- [InfoDeepSeek: Benchmarking Agentic Information Seeking for Retrieval-Augmented Generation](https://mlsanity.com/paper/2505.15872/) (2025)
- [Avoiding Over-Personalization with Rule-Guided Knowledge Graph Adaptation for LLM Recommendations](https://mlsanity.com/paper/2509.07133/) (2025)
- [Enhancing Serendipity Recommendation System by Constructing Dynamic User Knowledge Graphs with Large Language Models](https://mlsanity.com/paper/2508.04032/) (2025)
- [Generative Pseudo-Labeling for Pre-Ranking with LLMs](https://mlsanity.com/paper/2602.20995/) (2026)
- [Modular Representation Compression: Adapting LLMs for Efficient and Effective Recommendations](https://mlsanity.com/paper/2604.18146/) (2026)
- [SPADE: Escaping the Popularity-Similarity Frontier to Measure Serendipitous Recommendations](https://mlsanity.com/paper/2609.31164/) (2026)

## Abstract (arXiv)

Recommender systems (RSs) often suffer from the feedback loop phenomenon, e.g., RSs are trained on data biased by their recommendations. This leads to the filter bubble effect that reinforces homogeneous content and reduces user satisfaction. To this end, serendipity recommendations, which offer unexpected yet relevant items, are proposed. Recently, large language models (LLMs) have shown potential in serendipity prediction due to their extensive world knowledge and reasoning capabilities. However, they still face challenges in aligning serendipity judgments with human assessments, handling long user behavior sequences, and meeting the latency requirements of industrial RSs. To address these issues, we propose SERAL (Serendipity Recommendations with Aligned Large Language Models), a framework comprising three stages: (1) Cognition Profile Generation to compress user behavior into multi-level profiles; (2) SerenGPT Alignment to align serendipity judgments with human preferences using enriched training data; and (3) Nearline Adaptation to integrate SerenGPT into industrial RSs pipelines efficiently. Online experiments demonstrate that SERAL improves exposure ratio (PVR), clicks, and transactions of serendipitous items by 5.7%, 29.56%, and 27.6%, enhancing user experience without much impact on overall revenue. Now, it has been fully deployed in the "Guess What You Like" of the Taobao App homepage.