Listwise Preference Diffusion Optimization for User Behavior Trajectories Prediction

TL;DR

This paper introduces Listwise Preference Diffusion Optimization (LPDO), a diffusion-based framework for predicting multi-step user behavior trajectories by modeling global preferences, outperforming existing methods on real-world benchmarks.

Contribution

The paper proposes LPDO, a novel diffusion-based training method that explicitly models listwise preferences over entire sequences, addressing limitations of prior diffusion approaches.

Findings

LPDO outperforms state-of-the-art baselines on real-world benchmarks.

The method effectively captures global, listwise dependencies in user trajectories.

Proposes new metrics for evaluating multi-step sequence prediction quality.

Abstract

Forecasting multi-step user behavior trajectories requires reasoning over structured preferences across future actions, a challenge overlooked by traditional sequential recommendation. This problem is critical for applications such as personalized commerce and adaptive content delivery, where anticipating a user's complete action sequence enhances both satisfaction and business outcomes. We identify an essential limitation of existing paradigms: their inability to capture global, listwise dependencies among sequence items. To address this, we formulate User Behavior Trajectory Prediction (UBTP) as a new task setting that explicitly models long-term user preferences. We introduce Listwise Preference Diffusion Optimization (LPDO), a diffusion-based training framework that directly optimizes structured preferences over entire item sequences. LPDO incorporates a Plackett-Luce supervision signal and derives a tight variational lower bound aligned with listwise ranking likelihoods, enabling coherent preference generation across denoising steps and overcoming the independent-token assumption of prior diffusion methods. To rigorously evaluate multi-step prediction quality, we propose the task-specific metric Sequential Match (SeqMatch), which measures exact trajectory agreement, and adopt Perplexity (PPL), which assesses probabilistic fidelity. Extensive experiments on real-world user behavior benchmarks demonstrate that LPDO consistently outperforms state-of-the-art baselines, establishing a new benchmark for structured preference learning with diffusion models.