Pockettslm runs time-series health analysis fully on phones preserving privacy
PocketTSLM: Deploying an On-Device Soft-Prompt Time-Series Language Model for Privacy-Preserving Smartphone Reasoning
Machine LearningComputation and Language
Summary
Smartphones and wearables collect lots of health data over time, but usually treat it as simple summaries or text, losing important timing information. The paper’s authors created PocketTSLM, a way to run a special time-series language model directly on phones without sending sensitive data to the cloud. This method uses fewer data tokens and fits within phone memory limits, allowing analysis like sleep stage detection and ECG interpretation right on the device. This keeps personal health data private and reduces delays caused by offloading to cloud servers.
What this means in practice
- •For mobile app developers: Build apps that analyze physiological sensor signals locally on phones without sending private data to the cloud.
- •For wearable device engineers: Integrate privacy-preserving time-series analysis models that fit device constraints for health monitoring features.
Authors
Max Rosenblattl, Nikhil Krishnaswamy, Leon Nissen, Manuel Milling, Nicolas Zumarraga, Ning Wang, Thomas Kaar, Robert Jakob, Björn W. Schuller, Elgar Fleisch, Patrick Langer, Paul Schmiedmayer
Abstract
Consumer phones and wearables increasingly collect physiological time-series data, yet most mobile systems treat these streams as summaries or text rather than as temporal evidence. This representation is poorly matched to personal health systems that need to retrieve, route, or summarize sensor evidence while preserving temporal structure. Cloud-hosted time-series language models avoid smartphone limits, but they repeatedly move sensitive physiology off the device and add latency and per-query compute cost. Mobile model alternatives that aggregate time-series data into statistical features or serialize raw samples as text either compress temporal structure or exhaust the context and memory budgets of small on-device models. We present PocketTSLM, a mobile-native deployment of OpenTSLM SoftPrompt, the soft-prompt variant of the OpenTSLM time-series language model. PocketTSLM runs fully on a consumer smartphone by injecting encoded time-series windows as embeddings instead of tokenizing time series as text. A numerical parity protocol verifies identical greedy-decoded token sequences between the Swift port and a Python reference using Apple's MLX machine-learning framework. Answers also match the PyTorch reference, although rationale wording can differ. An on-device context-economics benchmark compares the soft-prompt path with a text-serialized baseline on the same backbone. In our benchmark format, soft-prompting uses approximately 12 times fewer time-series tokens and reduces prompt-prefill cost. PocketTSLM processes 30-second Sleep-EDF electroencephalography clips for sleep staging and a 1,000-sample (10 s) 12-lead electrocardiogram clip from ECG-QA. In comparison, text-serialized baselines exceed the phone's context or memory budget at longer time-series windows. All inference runs entirely on the device, with no health data transmitted during inference.