Researchers have introduced MuViS-C, a new benchmark designed to evaluate the robustness of learning-based virtual sensing systems against various sensor failures. The benchmark covers ten common failure modes across six domains and six different model architectures, including gradient-boosted trees, convolutional neural networks, and attention-based models. Findings indicate that all tested models degrade significantly under corruption, with gradient-boosted trees showing strong robustness, and dedicated robustification strategies improving attention-based models at the cost of nominal performance. The study emphasizes that model performance rankings vary across domains, highlighting the need for multi-domain evaluation. AI
IMPACT This benchmark will help researchers develop more reliable AI systems for critical cyber-physical applications by evaluating their performance under realistic sensor failure conditions.
RANK_REASON The cluster describes a new benchmark and research paper published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]
- arXiv
- attention
- CatalyzeX
- convolutional neural network
- DagsHub
- Gotit.pub
- Gradient Boosted Trees
- Hugging Face
- MLP-mixing
- MuViS-C
- ScienceCast
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
