HALO

IMU

Key facts

Parameters: ~35M trainable (+ frozen MiniLM 22.7M shared)
Venue: arXiv preprint (cs), 2026-08-27

Summary

Domain-specific IMU foundation model for open-vocabulary HAR across heterogeneous rates/channels/placements. Two-stage: heterogeneity-aware SSL encoder → synonym-aware soft contrastive alignment to frozen Sentence-BERT. ~35M trainable params; zero-shot = cosine retrieval over arbitrary label text (no per-dataset classifier).

Datasets used

Source paper

HALO: A Heterogeneity-Aware Language-Aligned IMU Foundation Model for Open-Set Human Activity Recognition