Sonata: A Hybrid World Model for Inertial Kinematics under Clinical Data Scarcity

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Delaney, Blaise, Patel, Salil, Xing, Yuji, Dootson, Dominic, Sevegnani, Karin, Antoniades, Chrystalina
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911639541907456
author Delaney, Blaise
Patel, Salil
Xing, Yuji
Dootson, Dominic
Sevegnani, Karin
Antoniades, Chrystalina
author_facet Delaney, Blaise
Patel, Salil
Xing, Yuji
Dootson, Dominic
Sevegnani, Karin
Antoniades, Chrystalina
contents We introduce Sonata, a compact latent world model for six-axis trunk IMU representation learning under clinical data scarcity. Clinical cohorts typically comprise tens to hundreds of patients, making web-scale masked-reconstruction objectives poorly matched to the problem. Sonata is a 3.77 M-parameter hybrid model, pre-trained on a harmonised corpus of nine public datasets (739 subjects, 190k windows) with a latent world-model objective that predicts future state rather than reconstructing raw sensor traces. In a controlled comparison against a matched autoregressive forecasting baseline (MAE) on the same backbone, Sonata yields consistently stronger frozen-probe clinical discrimination, prospective fall-risk prediction, and cross-cohort transfer across a 14-arm evaluation suite, while producing higher-rank, more structured latent representations. At 3.77 M parameters the model is compatible with on-device wearable inference, offering a step toward general kinematic world models for neurological assessment.
format Preprint
id arxiv_https___arxiv_org_abs_2604_18058
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Sonata: A Hybrid World Model for Inertial Kinematics under Clinical Data Scarcity
Delaney, Blaise
Patel, Salil
Xing, Yuji
Dootson, Dominic
Sevegnani, Karin
Antoniades, Chrystalina
Machine Learning
We introduce Sonata, a compact latent world model for six-axis trunk IMU representation learning under clinical data scarcity. Clinical cohorts typically comprise tens to hundreds of patients, making web-scale masked-reconstruction objectives poorly matched to the problem. Sonata is a 3.77 M-parameter hybrid model, pre-trained on a harmonised corpus of nine public datasets (739 subjects, 190k windows) with a latent world-model objective that predicts future state rather than reconstructing raw sensor traces. In a controlled comparison against a matched autoregressive forecasting baseline (MAE) on the same backbone, Sonata yields consistently stronger frozen-probe clinical discrimination, prospective fall-risk prediction, and cross-cohort transfer across a 14-arm evaluation suite, while producing higher-rank, more structured latent representations. At 3.77 M parameters the model is compatible with on-device wearable inference, offering a step toward general kinematic world models for neurological assessment.
title Sonata: A Hybrid World Model for Inertial Kinematics under Clinical Data Scarcity
topic Machine Learning
url https://arxiv.org/abs/2604.18058