NeuS 2025

2nd International Conference on
Neuro-symbolic Systems (NeuS)

May 28-30, 2025

University of Pennsylvania, Philadelphia, Pennsylvania, USA

Program

Note: Paper titles below link to pre-print versions of the papers. Final versions will be published after the conference.

Wednesday, May 28

08:00-09:00 Breakfast (AGH Lobby)
08:30-9:00 Welcome by Program Chairs
09:00-10:00 Keynote 1: Claire Tomlin
Session Chair: George Pappas
10:00-10:30 Coffee break
10:30-12:30 Research Paper Oral Presentations
Session Chair: Pradeep Ravikumar
  1. Why Neural Network Can Discover Symbolic Structures with Gradient-based Training: A Theoretic Foundation for Neurosymbolic AI with Practical Implications. Zhangyang Wang and Peihao Wang
  2. Provably Correct Automata Embeddings for Optimal Automata-Conditioned Reinforcement Learning. Beyazit Yalcinkaya, Niklas Lauffer, Marcell Vazquez-Chanlatte and Sanjit A. Seshia
  3. Stochastic Neural Simulation Relations for Transferring Control under Uncertainty. Alireza Nadali, Ashutosh Trivedi and Majid Zamani
  4. Learning Subject to Constraints via Abstract Gradient Descent. Shiwen Yu, Wanwei Liu, Zengyu Liu, Liqian Chen, Ting Wang, Naijun Zhan and Ji Wang
12:30-14:00 Lunch break
14:00-15:00 Keynote 2: Sriram Rajamani
Session Chair: Sanjit Seshia
15:00-15:30 Coffee break
15:30-17:00 Research Paper Oral Presentations
Session Chair: Rajeev Alur
  1. Neuro-symbolic Generative Diffusion Models for Physically Grounded, Robust, and Safe Generation. Jacob Christopher, Michael Cardei, Jinhao Liang and Ferdinando Fioretto
  2. Knowledge-Enriched Machine Learning for Tabular Data. Juyong Kim, Chandler Squires and Pradeep Ravikumar
  3. Differentiable Synthesis of Behavior Tree Architectures and Execution Nodes. Yu Huang, Ziji Wu, Kexin Ma and Ji Wang

Thursday, May 29

08:00-09:00 Breakfast (AGH Lobby)
09:00-10:30 Tutorial Paper Presentations
Session Chair: Sanjit Seshia
  1. On Supervised vs. Unsupervised Learning for First Order Hyperbolic Nonlinear PDEs: Applications to Traffic Modeling. Alexi Canesse, Zhe Fu, Nathan Lichtlé, Hossein Nick Zinat Matin, Zihe Liu, Maria Laura Delle Monache and Alexandre Bayen
  2. Efficient Processing of Neuro-Symbolic AI: A Tutorial and Case Study. Zishen Wan, Hanchen Yang, Ritik Raj, Che-Kai Liu, Arijit Raychowdhury and Tushar Krishna
  3. Specification-Guided Reinforcement Learning. Kishor Jothimurugan, Suguman Bansal, Osbert Bastani and Rajeev Alur
10:30-11:00 Coffee break
11:00-12:30 Research Paper Oral Presentations
Session Chair: Wenchao Li
  1. Logic Gate Neural Networks are Good for Verification. Fabian Kresse, Emily Yu, Christoph H. Lampert and Thomas A. Henzinger
  2. Learning Formal Specifications from Membership and Preference Queries. Ameesh Shah, Marcell Vazquez-Chanlatte, Sebastian Junges and Sanjit A. Seshia
  3. Learning Minimal Neural Specifications. Chuqin Geng, Zhaoyue Wang, Haolin Ye and Xujie Si
12:30-14:00 Lunch break
14:00-15:00 2 Minute Presentations For Each Poster
Session Chair: Pradeep Ravikumar
  1. Real-Time Reachability for Neurosymbolic Reinforcement Learning based Safe Autonomous Navigation. Nicholas Potteiger, Diego Manzanas-Lopez, Taylor T. Johnson and Xenofon Koutsoukos
  2. L*LM: Learning Automata from Demonstrations, Examples, and Natural Language. Marcell Vazquez-Chanlatte, Karim Elmaaroufi, Stefan Witwicki, Matei Zaharia and Sanjit A. Seshia
  3. Lean Copilot: Large Language Models as Copilots for Theorem Proving in Lean. Peiyang Song, Kaiyu Yang and Anima Anandkumar
  4. Neuro-Symbolic Discovery of Markov Population Processes. Luca Bortolussi, Francesca Cairoli, Julia Klein and Tatjana Petrov
  5. Taylor-Model Physics-Informed Neural Networks (PINNs) for Ordinary Differential Equations. Chandra Kanth Nagesh, Sriram Sankaranarayanan, Ramneet Kaur, Tuhin Sahai and Susmit Jha
  6. PCA-DDReach: Efficient Statistical Reachability Analysis of Stochastic Dynamical Systems via Principal Component Analysis. Navid Hashemi, Lars Lindemann and Jyotirmoy Deshmukh
  7. State-Dependent Conformal Perception Bounds for Neuro-Symbolic Verification of Autonomous Systems. Thomas Waite, Yuang Geng, Trevor S. Turnquist, Ivan Ruchkin and Radoslav Ivanov
  8. Neurosymbolic Finite and Pushdown Automata: Improved Multimodal Reasoning versus Vision Language Models (VLMs). Samuel Sasaki, Diego Manzanas Lopez and Taylor T. Johnson
  9. A Study of Modus Ponens in Transformer Models. Paulo Pirozelli and Fabio G. Cozman
  10. Modularity in Query-Based Concept Learning. Benjamin Caulfield and Sanjit A. Seshia
  11. Formal Synthesis of Lyapunov Stability Certificates for Linear Switched Systems using ReLU Neural Networks. Virginie Debauche, Alec Edwards, Raphaël M. Jungers and Alessandro Abate
  12. Taxonomic Networks: A Representation for Neuro-Symbolic Pairing. Zekun Wang, Ethan Haarer, Nicki Barari and Christopher MacLellan
  13. Four Principles for Physically Interpretable World Models. Jordan Peper, Zhenjiang Mao, Yuang Geng, Siyuan Pan and Ivan Ruchkin
  14. Automaton-Based Representations of Task Knowledge from Generative Language Models. Yunhao Yang, Cyrus Neary and Ufuk Topcu
  15. End-to-End Navigation with VLMs: Transforming Spatial Reasoning into Question-Answering. Dylan Goetting, Himanshu Singh and Antonio Loquercio
  16. ChatHTN: Interleaving Approximate (LLM) and Symbolic HTN Planning. Hector Munoz-Avila, David Aha and Paola Rizzo
  17. Mining Causal Signal Temporal Logic Formulas for Efficient Reinforcement Learning with Temporally Extended Tasks. Hadi Partovi Aria and Zhe Xu
  18. Expansion Span: Combining Fading Memory and Retrieval in Hybrid State Space Models. Elvis Nunez, Luca Zancato, Benjamin Bowman, Aditya Golatkar, Wei Xia and Stefano Soatto
  19. A Challenge to Build Neuro-Symbolic Video Agents. Sahil Shah, Harsh Goel, Sai Shankar Narasimhan, Minkyu Choi, S. P. Sharan, Oguzhan Akcin and Sandeep Chinchali
  20. KGAccel: A Domain-Specific Reconfigurable Accelerator for Knowledge Graph Reasoning. Hanning Chen, Ali Zakeri, Yang Ni, Fei Wen, Behnam Khaleghi, Hugo Latapie, Alvaro Velasquez and Mohsen Imani
  21. Neuro-Symbolic Behavior Trees and Their Verification. Serena Serbinowska, Diego Manzanas Lopez, Dung Thuy Nguyen and Taylor T. Johnson
  22. Observability of Latent States in Generative AI Models. Tian Yu Liu, Stefano Soatto, Matteo Marchi, Pratik Chaudhari and Paulo Tabuada
  23. Bidirectional End-to-End Framework for Transfer from Abstract Models in Non-Markovian Reinforcement Learning. Mahyar Alinejad, Precious Nwaorgu, Chinwendu Enyioha, Yue Wang, Alvaro Velasquez and George Atia
15:00-15:30 Coffee break
15:30-16:30 Poster session
16:30-17:00 Business Meeting
17:00-18:00
18:00-21:00 Banquet, Hall of Flags, Houston Hall, 3417 Spruce Street

Friday, May 30

08:00-09:00 Breakfast (AGH Lobby)
09:00-10:00 Panel Discussion:

Panelists: Zico Kolter, Armando Solar-Lezama, Alvaro Velasquez, and Shankar Sastry
Moderator: Pradeep Ravikumar
10:00-10:30 Coffee break
10:30-12:30 Research Paper Oral Presentations
Session Chair: Osbert Bastani
  1. From Road to Code: Neuro-Symbolic Program Synthesis for Autonomous Driving Scene Translation and Analysis. Johnathan Leung, Guansen Tong, Parasara Sridhar Duggirala and Praneeth Chakravarthula
  2. Assured Autonomy with Neuro-Symbolic Perception. R, Spencer Hallyburton and Miroslav Pajic
  3. Efficient Neuro-Symbolic Policy using In-Memory Computing. Tergel Molom-Ochir, Naman Saxena, Jiwoo Kim, Yiran Chen, Zhangyang Wang, Miroslav Pajic and Hai Li
  4. Interpretable Imitation Learning via Generative Adversarial STL Inference and Control. Wenliang Liu, Danyang Li, Erfan Aasi, Daniela Rus, Roberto Tron and Calin Belta
End of Conference

Keynote 1: Claire Tomlin

Title: Safe Learning for Robotics

Abstract: Automating safety-critical systems demands reliable, understandable control systems. Recent advances in neural networks for control present an exciting future if we can make guarantees about how the control behaves. In this talk, I will discuss three promising directions for neural networks in safety critical control: (1) using neural network to compute safety certificates, (2) certifying learned safety certificates, and (3) interpreting LLM tokens for robotic motion.

Bio: Claire Tomlin is a Professor of Electrical Engineering and Computer Sciences at UC Berkeley, where she holds the James and Katherine Lau Chair in Engineering. Her research interests include hybrid systems, distributed and decentralized optimization, and control theory, with an emphasis on applications, unmanned aerial vehicles, air traffic control and modeling of biological processes. She taught at Stanford University from 1998 to-2007 where she was a director of the Hybrid Systems Laboratory and held joint positions in the Department of Aeronautics and Astronautics and the Department of Electrical Engineering. She was awarded a MacArthur Genius grant in 2006 and the IEEE Transportation Technologies Award in 2017 "for contributions to air transportation systems, focusing on collision avoidance protocol design and avionics safety verification". She is a member of the National Academy of Engineering and the American Academy of Arts and Sciences.

Keynote 2: Sriram Rajamani

Title: Reimagining Large Scale Software Engineering with LLMs

Abstract: Over the past few years LLM-based-tools for code completion have taken the software engineering industry by storm. Tools like GitHub Copilot are used widely by engineers to improve programmer productivity. While LLMs offer significant advantages, several challenges arise with large code bases and large-scale software engineering problems. We believe that several of these challenges can be addressed by combining LLMs with techniques from static program analysis. We describe our work on building tools to solve large scale software engineering problems for very large code bases by combining LLMs together with static analysis methods and point to research opportunities in this area.

Bio: Sriram Rajamani is Corporate Vice President at Microsoft Research. Sriram is also an ACM fellow, INAE fellow, and winner of the Computer Aided Verification Award. His work has impacted both academic and industrial practice in programming languages, systems, security, and formal verification. He is currently working on reimagining the future of programming and software engineering in this era of large AI models. Sriram did his PhD in Computer Science at UC Berkeley.