publications

2026

1. ICML

   MultiBreak: A Scalable and Diverse Multi-turn Jailbreak Benchmark for Evaluating LLM Safety

   Jialin Song, X. Liu, W. Yang, and 4 more authors

   In International Conference on Machine Learning (ICML), 2026

   Abs arXiv

   A scalable multi-turn jailbreak benchmark built with active-learning-based data expansion across diverse harm categories, used to stress-test the safety behaviour of large language models and to expose persistent vulnerabilities to conversational attacks.

2. CVPR

   FluidGaussian: Propagating Simulation-Based Uncertainty Toward Functionally-Intelligent 3D Reconstruction

   Y. Liu^*, Jialin Song^*, M. Chanlatte, and 5 more authors

   In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

   Abs arXiv

   A physics-informed metric for evaluating next-best-view selection in 3D reconstruction, integrating geometric uncertainty through differentiable grid-based methods.

3. NatComm

   Learning earthquake ground motions via conditional generative modeling

   P. Ren, I. Naiman, M. Lacour, and 9 more authors

   Nature Communications, 2026

   Abs HTML

   A conditional generative model for simulating earthquake ground motion.

2025

1. ICML

   PDE-Controller: LLMs for Autoformalization and Reasoning of PDEs

   M. Soroco^*, Jialin Song^*, M. Xia, and 3 more authors

   In International Conference on Machine Learning (ICML), 2025

   Abs arXiv

   A large language model that converts natural-language descriptions of partial differential equations into formal temporal logic and executable control code, trained with a back-translation and auto-verification pipeline for data generation.

2. CPC

   Learning Spatiotemporal Dynamics from Sparse Data via a High-order Physics-encoded Network

   P. Ren^*, Jialin Song^*, C. Rao^*, and 4 more authors

   Computer Physics Communications, 2025

   Abs

   A high-order physics-encoded network for recovering spatiotemporal dynamics from sparse observations.

3. ACES

   HPC-Driven Modeling with ML-Based Surrogates for Magnon-Photon Dynamics in Hybrid Quantum Systems

   Jialin Song^*, Y. Tang^*, P. Ren, and 3 more authors

   In International Applied Computational Electromagnetics Society Symposium (ACES), 2025

   Abs arXiv

   A physics-informed deep learning surrogate for forecasting electric field dynamics, using curriculum learning to extend the prediction horizon and improve generalization.

2024

1. NeurIPS

   Data-Efficient Operator Learning via Unsupervised Pretraining and In-Context Learning

   W. Chen^*, Jialin Song^*, P. Ren, and 3 more authors

   In Advances in Neural Information Processing Systems (NeurIPS), 2024

   Abs arXiv

   An unsupervised pretraining and in-context learning pipeline that reduces the simulation data needed to train neural operators, studied across masking ratios, subsampling regimes and several PDE systems.

2. MLST

   Learning Spatiotemporal Dynamics From Sparse Measurements

   Jialin Song^*, Z. Song^*, P. Ren, and 3 more authors

   Machine Learning: Science and Technology, 2024

   Abs

   A neural ODE model for temporal forecasting from sparse measurements, paired with a Swin Transformer decoder that reconstructs the full spatiotemporal field, evaluated on turbulence and ERA5 climate data.