Sumit Kumar Jha
Eminent Scholar Chair Professor of Computer Science
384 CASE Building
11200 SW 8th Street
Miami, FL 33199
Prof. Jha’s research interests lie in the following areas:
- Artificial Intelligence
- Emerging Architectures
- Logic and Automata Theory
- Quantum Computing
Principal Investigator for $17 million in awarded projects since 2014, including DARPA GARD, AFRL ML-RCP, ONR Science of AI, Department of Energy EXPRESS, NSF Scalable Parallelism in the Extreme (SPX), NSF Software and Hardware Foundations (SHF), NSF Formal Methods in the Field (FMitF), NSF Secure and Trustworthy Computing (SaTC), National Nuclear Security Administration (NNSA), Department of Energy (DOE) ASCR RENEW, and AFOSR Young Investigator Program.
news
| Jan 23, 2025 | Our paper on metric-driven attributions in vision transformers accepted at ICLR 2025 |
|---|---|
| Jan 22, 2025 | Our paper on information flow in vision transformers accepted at AISTATS 2025 |
| Dec 15, 2024 | Our papers on jailbreaking LLMs and agent-based secure code generation presented in workshops @ NeurIPS 2024 |
| Sep 01, 2024 | Our paper on verification of path based computed presented at ICCAD 2024 |
| Feb 11, 2024 | Three (3) papers have been accepted at the ACM Design Automation Conference (DAC) 2024 |
selected publications
- Metric-Driven Attributions for Vision TransformersIn The Thirteenth International Conference on Learning Representations, ICLR, 2025
- Explaining ViTs Using Information FlowIn The 28th International Conference on Artificial Intelligence and Statistics, AISTATS, 2025
- Integrated Decision Gradients: Compute Your Attributions Where the Model Makes Its DecisionIn Thirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, February 20-27, 2024, Vancouver, Canada, 2024
- Jailbreaking Large Language Models with Symbolic MathematicsWorkshop on Socially Responsible Language Modelling Research (SoLaR) at NeurIPS, 2024
- AutoSafeCoder: A Multi-Agent Framework for Securing LLM Code Generation through Static Analysis and Fuzz TestingThe First Workshop on Safe & Trustworthy Agents (SATA) at NeurIPS, 2024
- On the Design of Novel Attention Mechanism for Enhanced Efficiency of Transformers61st ACM Design Automation Conference (DAC), 2024
- Zero-shot Detection of Out-of-Context Objects Using Foundation ModelsIn IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2025
- Exploring the Predictive Capabilities of AlphaFold Using Adversarial Protein SequencesIEEE Trans. Artif. Intell., 2024
- PATH: Evaluation of Boolean Logic Using Path-Based In-Memory Computing SystemsIEEE Trans. Comput. Aided Des. Integr. Circuits Syst., 2024
- Counterexample Guided Inductive Synthesis Using Large Language Models and Satisfiability SolvingIn IEEE Military Communications Conference, MILCOM 2023, Boston, MA, USA, October 30 - Nov. 3, 2023, 2023BEST PAPER CANDIDATE
- UpTime: Towards Flow-based In-Memory Computing with High Fault-ToleranceIn 60th ACM/IEEE Design Automation Conference, DAC 2023, San Francisco, CA, USA, July 9-13, 2023, 2023
- Automated Synthesis for In-Memory ComputingIn IEEE/ACM International Conference on Computer Aided Design, ICCAD 2023, San Francisco, CA, USA, October 28 - Nov. 2, 2023, 2023
- Shaping Noise for Robust Attributions in Neural Stochastic Differential EquationsIn Thirty-Sixth AAAI Conference on Artificial Intelligence, AAAI 2022, February 22 - March 1, 2022, 2022Oral Presentation
- ExplainIt!: A Tool for Computing Robust Attributions of DNNsIn Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, IJCAI 2022, Vienna, Austria, 23-29 July 2022, 2022
- Logic Synthesis for Digital In-Memory ComputingIn Proceedings of the 41st IEEE/ACM International Conference on Computer-Aided Design, ICCAD 2022, San Diego, California, USA, 30 October 2022 - 3 November 2022, 2022BEST PAPER CANDIDATE
- Towards resilient analog in-memory deep learning via data layout re-organizationIn DAC ’22: 59th ACM/IEEE Design Automation Conference, San Francisco, California, USA, July 10 - 14, 2022, 2022
- PATH: evaluation of boolean logic using path-based in-memory computingIn DAC ’22: 59th ACM/IEEE Design Automation Conference, San Francisco, California, USA, July 10 - 14, 2022, 2022
- On Smoother Attributions using Neural Stochastic Differential EquationsIn Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, IJCAI 2021, Virtual Event / Montreal, Canada, 19-27 August 2021, 2021
- COMPACT: Flow-Based Computing on Nanoscale Crossbars with Minimal SemiperimeterIn Design, Automation & Test in Europe Conference & Exhibition, DATE 2021, Grenoble, France, February 1-5, 2021, 2021BEST PAPER CANDIDATE
- Hybrid Analog-Digital In-Memory ComputingIn IEEE/ACM International Conference On Computer Aided Design, ICCAD 2021, Munich, Germany, November 1-4, 2021, 2021
- DP-MAP: Towards Resistive Dot-Product Engines with Improved PrecisionIn IEEE/ACM International Conference On Computer Aided Design, ICCAD 2020, San Diego, CA, USA, November 2-5, 2020, 2020
- Attribution-Based Confidence Metric For Deep Neural NetworksIn Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, 2019
- A Counterexample-Guided Approach to Parameter Synthesis for Linear Hybrid AutomataIn Hybrid Systems: Computation and Control, 11th International Workshop, HSCC 2008, St. Louis, MO, USA, April 22-24, 2008. Proceedings, 2008
- d-IRA: A Distributed Reachability Algorithm for Analysis of Linear Hybrid AutomataIn Hybrid Systems: Computation and Control, HSCC 2008, St. Louis, MO, USA, April 22-24, 2008. Proceedings, 2008
- Reachability for Linear Hybrid Automata Using Iterative Relaxation AbstractionIn Hybrid Systems: Computation and Control, HSCC 2007, Pisa, Italy, April 3-5, 2007, Proceedings, 2007
-