ARTFEED — Contemporary Art Intelligence

DeepFeature: LLM-Powered Feature Generation for Wearable Biosignals

publication · 2026-07-27

A research paper titled 'DeepFeature: LLM-Empowered Context-aware Feature Generation for Wearable Biosignals' has been published on arXiv. The paper proposes DeepFeature, a framework that uses large language models (LLMs) to generate context-aware features from wearable biosignals for healthcare applications. Existing feature extraction methods lack task-specific context, struggle with high-dimensional combinatorial spaces, and are prone to errors. DeepFeature integrates LLMs with expert knowledge and inter-feature interactions, employing iterative refinement based on feature assessment feedback. The paper is available at https://arxiv.org/abs/2512.08379.

Key facts

  • Paper titled 'DeepFeature: LLM-Empowered Context-aware Feature Generation for Wearable Biosignals'
  • Published on arXiv with ID 2512.08379
  • Proposes DeepFeature framework for wearable biosignal feature generation
  • Uses large language models (LLMs) for context-aware feature extraction
  • Addresses limitations of existing methods: lack of context, high-dimensional space, errors
  • Integrates multi-source generation: LLMs, expert knowledge, inter-feature interactions
  • Employs iterative refinement with feature assessment feedback
  • Targets healthcare applications using wearable devices

Entities

Institutions

  • arXiv

Sources