TRACE-TS: Attribution-Grounded and Traceable Sensor-Language Reasoning for Human Activity Understanding

Wearable sensors capture fine-grained motion patterns that support rich behavioral understanding, yet most existing methods reduce these signals to activity labels. Recent LM-based approaches generate natural-language explanations for sensor data, but their reasoning is weakly grounded in the underlying signal, leading to fluent yet unverifiable explanations. We introduce \textsc{TRACE-TS} (Traceable Reasoning with Attribution-Grounded Evidence), a framework for structured and signal-grounded reasoning over wearable time series. \textsc{TRACE-TS} uses attribution from an expert classifier to identify salient spatio-temporal sensor regions, uses them to construct DAG reasoning traces with explicit evidence provenance, and trains a compact language model to generate these traces through gated cross-attention over sensor memory tokens. At inference, the adapted model jointly outputs the activity prediction and its reasoning trace, without requiring attribution computation or teacher guidance. We introduce \emph{Semantic Node Match} (SNM), an LLM-as-judge metric that diagnoses reasoning fidelity at the observation, inference, and synthesis levels, localizing hallucinated observations and broken evidence chains missed by standard NLG metrics. Across seven wearable benchmarks, \textsc{TRACE-TS} achieves the best average accuracy and F1 among all evaluated methods (\textbf{84.43%}/\textbf{81.24%}), and outperforms the best LLM-based baseline by \textbf{17.96}% in F1. Our code is available at {https://anonymous.4open.science/r/TRACE-TS-F0A4}.