Hamiltonian Dynamics Model Captures User Preference Inertia

_reachsumit · x · 2026-08-27

This paper introduces HSR, a sequential recommender that models preference evolution as a dissipative Hamiltonian system. By decomposing preferences into position (stable) and momentum (short-term), it captures inertia, periodicity, and shifts that first-order models miss. Experiments show HSR outperforms Transformer and SSM-based models on benchmarks.

Original post →

More from Research

Research channel →