Gagandeep Singh
I am an Assistant Professor in the Department of Computer Science at the University of Illinois Urbana-Champaign (UIUC) where I lead the FOrmally Certified Automation and Learning (FOCAL) Lab. My current focus is on combining ideas from Formal Logic, Machine Learning, and Systems research to develop systematic and principled approaches (as opposed to “IT WORKS”) for constructing intelligent compute systems with formal guarantees about their behavior and safety. Our group has received the NSF Career Award, Google Research Scholar Award, and Qualcomm Innovation Fellowship. Some of the topics (in alphabetical order) that we currently work on are
- Deep Learning for Systems
- Reinforcement Learning from Human Feedback
- Explainable AI
- Federated Learning
- Logic-guided Learning
- Machine Learning for Wireless
- Neural Network Verification
- NeuroSymbolic Program Analysis
- Out-of-Distribution Generalization
- Safety of Large Language Models
- Static Analysis of Differentiable Programs
I obtained a PhD in Computer Science from ETH Zurich in 2020 working with Prof. Markus Püschel and Prof. Martin Vechev. During my PhD, I designed scalable and precise automated reasoning methods and tools for programs and deep neural networks. I co-received the ACM SIGPLAN Doctoral Dissertation Award given annually to the best dissertations in the area of Programming Languages. Before that, I completed a Masters in Computer Science at ETH in 2014 receiving the ETH Master Medal and Bachelors in Computer Science and Engineering from IIT Patna in 2012 receiving the President of India Gold Medal.
For prospective PhD candidates: Please apply to the UIUC CS graduate program and mention my name. I will get back to you if there is a good match.
For UIUC students looking for a paid RA: We currently do not have any openings.
News
- 26 Sep 2024: 1 paper accepted at NeurIPS'24 Datasets and Benchmarks Track.
- 25 Sep 2024: Two papers accepted at NeurIPS'24.
- 1 July 2024: Our work on building the first certified training method against UAP attack is conditionally accepted at ECCV'24.
- 19 June 2024: We have released the ORL framework for preference-based offline RL, which forms the basis of RLHF.
- 14 June 2024: Checkout the exciting Master thesis work by Shaurya Gomber on Neural Abstract Interpretation
- 30 May 2024: We have released the first framework for certifying bias in the output of LLMs . The code is avaialble on Github . A preprint of the work is available on Arxiv
- 1 May 2024: 2 papers accepted at ICML'24
- 31 March 2024: 1 paper accepted at PLDI'24
- 27 March 2024: We have released the first DSL called ConstraintFlow for Neural Network Certification.
- 5 March 2024: We have released a general framework for grammar-guided code generation with LLMs. The code is avaialble on Github . A preprint of the work is available on Arxiv
- 27 Feb 2024: We released the first framework on quantifying knowledge comprehension in LLMs on Arxiv
- 23 Feb 2024: 1 paper accepted at MLSys'24
- 16 Jan 2024: 3 papers accepted at ICLR'24
- 1 Jan 2024: Our latest work on designing an efficient learning algorithm for offline preference-based deep reinforcement learning is available on Arxiv .
- 20 Dec 2023: Our latest work develops priming attacks to bypass the safety training of open-source LLMs. More details are available at https://llmpriming.focallab.org/
- Nov 3 2023: Excited to receive a grant from the Amazon Illinois Center on Artificial Intelligence for Interactive Conversational Experiences on Fairness Verification of LLMs .
- 23 Oct 2023: Gave a keynote at SAS 2023 on Trustworthy AI
- 13 July 2023: Our paper on creating reward poisoning attacks on deep reinforcement learning is accepted at TMLR with featured certification (Oral/Spotlight category at TMLR)
- 11 July 2023: Our paper on interpreting robust proofs of neural networks won an outstanding paper award at WFVML@ICML'23
- 30 June 2023: Our paper on synthesizing static analyzers for AutoDiff code is accepted at OOPSLA'23
- 19 April 2023: Excited to receive the Google Research Scholar Award
- 4 April 2023: Our paper on incremental verification of neural networks is accepted at PLDI'23
- 3 Feb 2023: Delighted to receive the NSF Career Award for our proposal on building incremental neural network verifiers.
- 20 Jan 2023: Our paper on training neural networks provably robust against geometric perturbations accepted for spotlight presentation at ICLR'23
- 9 Dec 2022: Our paper on creating real-world hardware attacks on ML-based wireless systems accepted at NSDI'23
- 9 Nov 2022: Honored to be serving on the ACM India Doctoral Dissertation Award Committee
- 2 Sep 2022: Our papers on verification of GNN-based schedulers and Abstract Interpretation of Higher-Order Automatic Differentiation accepted in round 2 of OOPSLA'22
- 30 Aug 2022: Congratulations to Calvin for winning the prestigious Qualcomm Innovation Fellowship 2022
- 1 Aug 2022: Our proposal Provably Robust Machine Learning for Next Generation Cellular Networks has been selected for funding under the NSF RINGS program
- 1 May 2022: Our paper on proof sharing for neural network verificationaccepted at CAV'22
- 25 Feb 2022: Our paper on proof transfer for neural network verificationaccepted in round 1 of OOPSLA'22
- 20 Jan 2022: Our paper on provably robust adversarial example generation accepted at ICLR'22
- 27 Sep 2021: Our papers on static analysis of Clarke Jacobians and scaling multi-neuron verification of neural networks accepted at POPL'22