cs.AINov 10, 2026

SeqFold: Timestamp-Addressable Compressed States for User Action Sequence Backbones

Authors: Mianwei Zhou, Duo Zhang, Henry Tsai, Yuqi Zheng, Jian Li, Juexin Lin, Sujay Khandagale, Vijay Rajaram, 

Organizations: Coupang Mountain View, United States

Abstract

Existing sequential user modeling methods mainly focus on improving the backbone encoder or learning task-specific user representations for a single prediction objective. We instead study how to export the multi-layer user states of a frozen long-history backbone as compact, timestamp-addressable representations for downstream training backfill and serving. In this setting, the user state at a given timestamp summarizes the user's actions before that time, so downstream systems need only the state aligned to each target timestamp rather than the full hidden-state timeline. To this end, we present SeqFold, a compressed-state interface that combines learned multi-channel fusion, channel-specific temporal decay, and Lookup-Free Quantization (LFQ) to produce compact discrete codes from frozen backbone states. Task-specific decoders then map these shared codes into flexible embedding spaces for downstream models. To make timestamp-addressable export practical at scale, we further develop a fused streaming implementation that updates the fusion recurrence after every action while materializing states only at target timestamps. On large-scale proprietary datasets, SeqFold reduces per-state representation storage by 3,400×3{,}400\times while preserving most of the downstream ranking performance of the uncompressed model, showing that SeqFold is an effective interface for reusing long-history sequence backbones in downstream models.

Figures & tables

Explore similar work

CardsList