Zishen Wan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zishen Wan has a total of 115 co-authors. Here are the top co-authors by number of shared publications:
- Arijit Raychowdhury
Georgia Institute of Technology
11 shared publications - Tushar Krishna
Georgia Institute of Technology, Atlanta, GA, USA
7 shared publications - Mohamed Ibrahim
Georgia Institute of Technology, Atlanta, GA, USA
5 shared publications - Vijay Janapa Reddi
Harvard University
5 shared publications - Che-Kai Liu
Georgia Institute of Technology
4 shared publications - Yang (Katie) Zhao
University of Minnesota, Twin Cities, Minneapolis, MN, USA
3 shared publications
Latest Publications
- Open Access
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Faster-MoA: Low-Latency Tree-Structured MoA Serving with Early Exit and Agent-Aware Prefill-Decode Overlap
- 26 July 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessFaster-MoA: Low-Latency Tree-Structured MoA Serving with Early Exit and Agent-Aware Prefill-Decode Overlap
- Open Access
- 8 July 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 25 (4) , 1-44
Cross-Layer Design of Vector-Symbolic Computing: Bridging Cognition and Brain-Inspired Hardware Acceleration
Vector Symbolic Architectures (VSAs), also known as hyperdimensional (HD) computing, are increasingly deployed in cognitive applications due to their simple and efficient operations. The widespread adoption has, in turn, spurred the development of a diverse set of hardware solutions that optimize VSA performance for embedded and edge AI systems. Despite these advances, there remains a lack of comprehensive, unified discussion on the co-design and co-evolution of VSA algorithms and hardware. This survey aims at bridging that gap by linking theoretical, software-level explorations with efficient hardware architectures and emerging technology fabrics for VSAs, providing co-design insights that are accessible to both algorithm and hardware communities. First, we introduce the principles of vector-symbolic computing, including its core mathematical operations and learning paradigms. Second, we provide an in-depth discussion on hardware technologies for VSAs, analyzing analog, mixed-signal, and digital circuit design styles. We compare hardware implementations of VSAs by carrying out detailed analysis of their performance characteristics and tradeoffs, from which we distill design guidelines that are applicable across arbitrary VSA formulations. Third, we discuss a methodology for cross-layer design of VSAs that identifies synergies across layers and explores key ingredients for hardware/software co-design of VSAs. Finally, as a concrete case study of this methodology, we present an in-memory computing hardware design for VSA-based hierarchical cognition, illustrating how the proposed co-design principles translate into efficient architectures. The article concludes with a discussion of open research challenges and opportunities for future explorations.
- 8 July 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Embedded Computing Systems
- Vol. 25 (4) , 1-44
Open AccessCross-Layer Design of Vector-Symbolic Computing: Bridging Cognition and Brain-Inspired Hardware Acceleration
Vector Symbolic Architectures (VSAs), also known as hyperdimensional (HD) computing, are increasingly deployed in cognitive applications due to their simple and efficient operations. The widespread adoption has, in turn, spurred the development of a diverse set of hardware solutions that optimize VSA performance for embedded and edge AI systems. Despite these advances, there remains a lack of comprehensive, unified discussion on the co-design and co-evolution of VSA algorithms and hardware. This survey aims at bridging that gap by linking theoretical, software-level explorations with efficient hardware architectures and emerging technology fabrics for VSAs, providing co-design insights that are accessible to both algorithm and hardware communities. First, we introduce the principles of vector-symbolic computing, including its core mathematical operations and learning paradigms. Second, we provide an in-depth discussion on hardware technologies for VSAs, analyzing analog, mixed-signal, and digital circuit design styles. We compare hardware implementations of VSAs by carrying out detailed analysis of their performance characteristics and tradeoffs, from which we distill design guidelines that are applicable across arbitrary VSA formulations. Third, we discuss a methodology for cross-layer design of VSAs that identifies synergies across layers and explores key ingredients for hardware/software co-design of VSAs. Finally, as a concrete case study of this methodology, we present an in-memory computing hardware design for VSA-based hierarchical cognition, illustrating how the proposed co-design principles translate into efficient architectures. The article concludes with a discussion of open research challenges and opportunities for future explorations.
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 46 (4) , 134-143
Architecture 2.0: When Artificial Intelligence Starts Designing the Chips: A Fireside Chat With Vijay Janapa Reddi
- 1 July 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 46 (4) , 134-143
Architecture 2.0: When Artificial Intelligence Starts Designing the Chips: A Fireside Chat With Vijay Janapa Reddi
- Open Access
- 22 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 493-510
CREATE: C ross-Layer Resilience Characterization and Optimization for Efficient yet R eliable E mbodied A I Sys te ms
Embodied Artificial Intelligence (AI) has recently attracted significant attention as it bridges AI with the physical world. Modern embodied AI systems often combine a Large Language Model (LLM)-based planner for high-level task planning and a reinforcement learning (RL)-based controller for low-level action generation, enabling embodied agents to tackle complex tasks in real-world environments. However, deploying embodied agents remains challenging due to their high computation requirements, especially for battery-powered local devices. Although techniques like lowering operating voltage can improve energy efficiency, they can introduce bit errors and result in task failures. In this work, we propose CREATE, a general design principle that leverages heterogeneous resilience at different layers for synergistic energy-reliability co-optimization. For the first time, we conduct a comprehensive error injection study on modern embodied AI systems and observe an inherent but heterogeneous fault tolerance. Building upon these insights, we develop an anomaly detection and clearance mechanism at the circuit level to eliminate outlier errors. At the model level, we propose a weight-rotation-enhanced planning algorithm to improve the fault tolerance of the LLM-based planner. Furthermore, we introduce an application-level technique, autonomy-adaptive voltage scaling, to dynamically adjust the operating voltage of the controllers. The voltage scaling circuit is co-designed to enable online voltage adjustment. Extensive experiments demonstrate that without compromising task quality, CREATE achieves 40.6% computational energy savings on average over nominal-voltage baselines and 35.0% over prior-art techniques. This further leads to 29.5% to 37.3% chip-level energy savings and approximately a 15% to 30% improvement in battery life.
- 22 March 2026
- Published by Association for Computing Machinery (ACM)
- p. 493-510
Open AccessCREATE: C ross-Layer Resilience Characterization and Optimization for Efficient yet R eliable E mbodied A I Sys te ms
Embodied Artificial Intelligence (AI) has recently attracted significant attention as it bridges AI with the physical world. Modern embodied AI systems often combine a Large Language Model (LLM)-based planner for high-level task planning and a reinforcement learning (RL)-based controller for low-level action generation, enabling embodied agents to tackle complex tasks in real-world environments. However, deploying embodied agents remains challenging due to their high computation requirements, especially for battery-powered local devices. Although techniques like lowering operating voltage can improve energy efficiency, they can introduce bit errors and result in task failures. In this work, we propose CREATE, a general design principle that leverages heterogeneous resilience at different layers for synergistic energy-reliability co-optimization. For the first time, we conduct a comprehensive error injection study on modern embodied AI systems and observe an inherent but heterogeneous fault tolerance. Building upon these insights, we develop an anomaly detection and clearance mechanism at the circuit level to eliminate outlier errors. At the model level, we propose a weight-rotation-enhanced planning algorithm to improve the fault tolerance of the LLM-based planner. Furthermore, we introduce an application-level technique, autonomy-adaptive voltage scaling, to dynamically adjust the operating voltage of the controllers. The voltage scaling circuit is co-designed to enable online voltage adjustment. Extensive experiments demonstrate that without compromising task quality, CREATE achieves 40.6% computational energy savings on average over nominal-voltage baselines and 35.0% over prior-art techniques. This further leads to 29.5% to 37.3% chip-level energy savings and approximately a 15% to 30% improvement in battery life.
- 10 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-16
A 40-nm Programmable Heterogeneous SoC With 5.625/0.85 MB RRAM/SRAM for Accelerating Neuro-Symbolic AI Models
Neuro-symbolic (NeSy) artificial intelligence (AI) integrates neural learning with symbolic reasoning to enable data-efficient, interpretable, and generalizable intelligence, making it a promising paradigm for human-like cognition. However, the heterogeneous and dynamic execution patterns of NeSy models pose fundamental challenges to conventional AI hardware, which is typically optimized for dense matrix operations. This article presents a fully programmable heterogeneous system-on-chip (SoC) fabricated in 40-nm CMOS, specifically designed to accelerate a wide range of NeSy workloads. The architecture features: 1) integrated resistive RAM (RRAM) and static random access memory (SRAM) neural-symbolic data paths; 2) ultra-dense (4.80 Mb/mm2), energy-efficient (0.247 pJ/b) RRAM macros with edge-triggered and prolonged sensing; 3) scheduler-informed power management; and 4) programming support for variable resolution, vector lengths, and batching. The SoC supports multiple NeSy paradigms and achieves up to 10.8 TOPS/W energy efficiency and $6.47\times $ power reduction across reasoning benchmarks. This work demonstrates the first end-to-end silicon system for generalizable and efficient NeSy inference.
- 10 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Solid-State Circuits
- Vol. PP (00189200) , 1-16
A 40-nm Programmable Heterogeneous SoC With 5.625/0.85 MB RRAM/SRAM for Accelerating Neuro-Symbolic AI Models
Neuro-symbolic (NeSy) artificial intelligence (AI) integrates neural learning with symbolic reasoning to enable data-efficient, interpretable, and generalizable intelligence, making it a promising paradigm for human-like cognition. However, the heterogeneous and dynamic execution patterns of NeSy models pose fundamental challenges to conventional AI hardware, which is typically optimized for dense matrix operations. This article presents a fully programmable heterogeneous system-on-chip (SoC) fabricated in 40-nm CMOS, specifically designed to accelerate a wide range of NeSy workloads. The architecture features: 1) integrated resistive RAM (RRAM) and static random access memory (SRAM) neural-symbolic data paths; 2) ultra-dense (4.80 Mb/mm2), energy-efficient (0.247 pJ/b) RRAM macros with edge-triggered and prolonged sensing; 3) scheduler-informed power management; and 4) programming support for variable resolution, vector lengths, and batching. The SoC supports multiple NeSy paradigms and achieves up to 10.8 TOPS/W energy efficiency and $6.47\times $ power reduction across reasoning benchmarks. This work demonstrates the first end-to-end silicon system for generalizable and efficient NeSy inference.
- Open Access
- 11 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 67-84
Compositional AI Beyond LLMs: System Implications of Neuro-Symbolic-Probabilistic Architectures
- 11 December 2025
- Published by Association for Computing Machinery (ACM)
- p. 67-84
Open AccessCompositional AI Beyond LLMs: System Implications of Neuro-Symbolic-Probabilistic Architectures
- Open Access
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1838-1851
RTGS: Real-Time 3D Gaussian Splatting SLAM via Multi-Level Redundancy Reduction
- 17 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1838-1851
Open AccessRTGS: Real-Time 3D Gaussian Splatting SLAM via Multi-Level Redundancy Reduction
- 4 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 45 (3) , 103-107
Sipping Matcha of Security: A Fireside Chat With Mengjia Yan
- 4 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Micro
- Vol. 45 (3) , 103-107
Sipping Matcha of Security: A Fireside Chat With Mengjia Yan
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 982-997
ReCA: Integrated Acceleration for Real-Time and Efficient Cooperative Embodied Autonomous Agents
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 982-997
Open AccessReCA: Integrated Acceleration for Real-Time and Efficient Cooperative Embodied Autonomous Agents
- Open Access
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 704-718
OctoCache: Caching Voxels for Accelerating 3D Occupancy Mapping in Autonomous Systems
- 30 March 2025
- Published by Association for Computing Machinery (ACM)
- p. 704-718
Open AccessOctoCache: Caching Voxels for Accelerating 3D Occupancy Mapping in Autonomous Systems