paper-with-me

Papers

Gravity-Aware Hierarchical Routing for Lightweight SensorLLM on Human Activity Recognition

2026-06-01 · Hao Li, Mingrui Zheng, Yasuyuki Tahara, Yuichi Sei arxiv

Recent studies on sensor-language alignment have shown that two-stage frameworks can improve the semantic modeling ability of wearable-sensor human activity recognition (HAR), where SensorLLM-style methods first perform motion-to-language alignment and then fine-tune the model for downstream tasks. However, our experiments reveal a consistent failure mode when the Stage 2 backbone is compressed to a compact model such as TinyLlama: recognition of dynamic activities remains relatively strong, while the discrimination of low-motion static classes such as standing, sitting, and lying degrades substantially. To address this issue, we propose a gravity-aware hierarchical routing head as a lightweight post-alignment adaptation built on top of an already aligned model, rather than a new large-scale pretraining framework. The method uses the per-channel mean and std from the Chronos tokenizer state to extract statistical cues related to posture and gravity direction, and adaptively combines a static expert and a full expert through soft routing, together with a load-balancing loss for stable training. On the MHealth dataset, this design significantly improves macro-F1 with minimal parameter overhead, and the gains are concentrated mainly on static classes while preserving strong performance on dynamic activities. As a first arXiv disclosure, the current paper reports results on a single dataset only, with the goal of highlighting the core method and laying the groundwork for broader evaluation in future work.

📄 PDF Abstract BibTeX arXiv:2606.04019

Code (0)

등록된 구현이 없습니다.

Tasks

Human Activity Recognition

Similar Papers 제목 키워드 기반

SensorLLM: Human-Intuitive Alignment of Multivariate Sensor Data with LLMs for Activity Recognition

2024-10-14 · Zechen Li, Shohreh Deldari, Linyao Chen, Hao Xue 외

We introduce SensorLLM, a two-stage framework that enables Large Language Models (LLMs) to perform human activity recognition (HAR) from wearable sensor data. While LLMs excel at reasoning and generalization, they strugg…

Activity RecognitionDescriptiveHuman Activity RecognitionQuestion Answering+2

Gravity-Awareness: Deep Learning Models and LLM Simulation of Human Awareness in Altered Gravity

2025-10-29 · Bakytzhan Alibekov, Alina Gutoreva, Elisa Raffaella-Ferre arxiv

Earth s gravity fundamentally shapes human behaviour. The brain encodes this force as an internal model of gravity, enabling the prediction and interpretation of gravitational effects during perception and action. Unders…

HierRouter: Coordinated Routing of Specialized Large Language Models via Reinforcement Learning

2025-11-13 · Nikunj Gupta, Bill Guo, Rajgopal Kannan, Viktor K. Prasanna arxiv

Large Language Models (LLMs) deliver state-of-the-art performance across many tasks but impose high computational and memory costs, limiting their deployment in resource-constrained or real-time settings. To address this…

Reinforcement LearningMathematical ReasoningCode Generation

A Novel Hierarchical Ant based QoS aware Intelligent Routing Scheme for MANETS

2013-12-07 · Debajit Sensarma, Koushik Majumder

MANET is a collection of mobile devices with no centralized control and no pre-existing infrastructures. Due to the nodal mobility, supporting QoS during routing in this type of networks is a very challenging task. To ta…

ClusteringManagement

CARGO: A Framework for Confidence-Aware Routing of Large Language Models

2025-09-18 · Amine Barrak, Yosr Fourati, Michael Olchawa, Emna Ksontini 외 arxiv

As large language models (LLMs) proliferate in scale, specialization, and latency profiles, the challenge of routing user prompts to the most appropriate model has become increasingly critical for balancing performance a…