Ziyao Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Ziyao Wang has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Ang Li
University of Maryland College Park
2 shared publications - Yexiao He
Department of Electrical and Computer Engineering, University of Maryland, College Park, 8223 Paint Branch Drive, College Park, 20742, MD, USA
2 shared publications Adam Porter
Department of Computer Science, University of Maryland, College Park, 8125 Paint Branch Dr, College Park, 20742, MD, USA
1 shared publication- Guoheng Sun
University of Maryland College Park, College Park, MD, USA
1 shared publication Hongli Wang
Department of Chemistry and Biochemistry, University of Notre Dame, Holy Cross Dr, Notre Dame, 46556, IN, USA
1 shared publication- Isabel Thompson
Department of Chemistry and Biochemistry, University of Notre Dame, Holy Cross Dr, Notre Dame, 46556, IN, USA
1 shared publication
Latest Publications
- Open Access
LIMBER: A Hybrid RNA Foundation Model for Predicting and Understanding RNA Structure and Function
RNA foundation models offer a promising route to decoding the relationships between RNA sequence, structure and function, yet how model architecture shapes the acquisition and organization of biological information remains poorly understood. Here we introduce Layer-Interleaved Mamba-Bidirectional Encoder for RNA (LIMBER), a hybrid foundation model integrating bidirectional self-attention with selective state-space modeling, pre-trained on 40.5 million non-coding RNA sequences. With task-specific fine-tuning, LIMBER achieves leading performance on secondary-structure prediction and non-coding RNA classification benchmarks, while demonstrating strong transferability to translation-related tasks and RNA–small-molecule interaction prediction. Beyond model development, we establish a unified, task-agnostic framework to systematically evaluate and interpret the biological knowledge intrinsically acquired by RNA foundation models. This framework validates LIMBER’s strong capabilities in recovering RNA structures and predicting mutational effects, while uncovering a hierarchical, task-dependent organization of biological information across model layers. In addition, we demonstrate distinct scaling behaviors across structural, evolutionary, and functional capabilities by training and evaluating a broad panel of model variants spanning model capacity, training duration, and dataset size. Taken together, these findings establish LIMBER as a versatile RNA foundation model and provide insight into how architectural design, internal representations and scaling shape the learning of RNA sequence-structure-function relationships.
Open AccessLIMBER: A Hybrid RNA Foundation Model for Predicting and Understanding RNA Structure and Function
RNA foundation models offer a promising route to decoding the relationships between RNA sequence, structure and function, yet how model architecture shapes the acquisition and organization of biological information remains poorly understood. Here we introduce Layer-Interleaved Mamba-Bidirectional Encoder for RNA (LIMBER), a hybrid foundation model integrating bidirectional self-attention with selective state-space modeling, pre-trained on 40.5 million non-coding RNA sequences. With task-specific fine-tuning, LIMBER achieves leading performance on secondary-structure prediction and non-coding RNA classification benchmarks, while demonstrating strong transferability to translation-related tasks and RNA–small-molecule interaction prediction. Beyond model development, we establish a unified, task-agnostic framework to systematically evaluate and interpret the biological knowledge intrinsically acquired by RNA foundation models. This framework validates LIMBER’s strong capabilities in recovering RNA structures and predicting mutational effects, while uncovering a hierarchical, task-dependent organization of biological information across model layers. In addition, we demonstrate distinct scaling behaviors across structural, evolutionary, and functional capabilities by training and evaluating a broad panel of model variants spanning model capacity, training duration, and dataset size. Taken together, these findings establish LIMBER as a versatile RNA foundation model and provide insight into how architectural design, internal representations and scaling shape the learning of RNA sequence-structure-function relationships.
- Open Access
- 23 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 138-153
EdgeLoRA: An Efficient Multi-Tenant LLM Serving System on Edge Devices
- 23 June 2025
- Published by Association for Computing Machinery (ACM)
- p. 138-153
Open AccessEdgeLoRA: An Efficient Multi-Tenant LLM Serving System on Edge Devices