DeepDOSReasoner: Physics-Grounded Reasoning for Density of States Prediction
Nature Computational Science 2026
(Under Review)
Yingheng Wang
PhD Candidate, Computer Science · Cornell University
About
I am a final-year PhD candidate in Computer Science at Cornell University with Carla P. Gomes. I also work closely with Volodymyr Kuleshov and Christopher De Sa. I am affiliated with Cornell AI for Science Institute and AI-LEAF Institute. My research studies knowledge-centric AI with an emphasis on scientific reasoning. I was also student researcher at Amazon Grand Challenge, AWS AI Lab, Microsoft Research, and NEC Lab America, where my research led to impactful papers and patents on computational antibody design, protein language models, time-series foundation models, and efficient neural architectures. My research has been generously supported by Cornell Presidential Life Science Fellowship.
I am currently on the job market — please feel free to reach out!
Research Interests
My research centers on developing knowledge-centric AI methods for scientific reasoning, which integrate physical and domain knowledge into modern AI techniques to improve their efficacy, efficiency, robustness, and interpretability. Current research topics include:
- Learning with physical knowledge: constrained learning ICML26, symmetry learning AAAI24, disentangled representations ICML23
- Learning under limited resources: self-supervised & unsupervised learning AAAI23 TheWebConf, multi-modal learning ICLR25 AI4Mat expressive and efficient neural networks ICLR25 NeurIPS23 ICML22
- Reasoning & inference with learning: reinforcement learning ICLR26 ICLR26, constrained reasoning ICLR26 AAAI26 AAAI26, probabilistic inference AAAI25
- Generative AI: large language models ICLR25 ICML26, diffusion & flow-based models ICML23 AAAI25 ICCV25 CV4MS, generative agents ICLR26
- Applications: scientific discovery (biology, materials, medicine, and sustainability)
- Taxonomy & Datasets & Benchmarks: genrative molecule design Nature Machine Intelligence, machine learning for materials discovery NeurIPS23, structured data augmentation IEEE Data Engineering Bulletin, expert-level combinatorial reasoning of LLMs ICLR26, expert-level scientific reasoning of LLMs Nature
Selected Publications
AbMPO: Multi-Objective Policy Optimization for Affinity-Calibrated Antibody Sequence Design
NeurIPS 2026
(Under Review)
Deep Scientific Reasoning under Physical Constraints: Structure-Aware Spectrum Prediction
ICML 2026
HeuriGym: An Agentic Benchmark for LLM-Crafted Heuristics in Combinatorial Optimization
ICLR 2026
LC-PLM: Long-context Protein Language Modeling Using Bidirectional Mamba with Shared Projection Layers
ICLR 2026
🏆 TMLR Featured Certificate
InfoDiffusion: Representation Learning Using Information Maximizing Diffusion Models
ICML 2023
Services
Area Chair: NeurIPS (AI4Science), ICML (AI4Science).
Conference Reviewer: ICML, NeurIPS, ICLR, AAAI, AISTATS, UAI, CVPR, ICCV, ECCV, COLM, KDD.
Journal Reviewer: IEEE Transactions on Knowledge and Data Engineering (TKDE), Transactions on Machine Learning Research (TMLR), Pattern Recognition, Bioinformatics, Journal of Biomedical Informatics (JBI).
Workshop Reviewer/PC: NeurIPS (Math-AI, AI4Science, GenBio, GLFrontiers), ICLR (AGI, AI4Science), ICML (AI4Science), KDD (Health Day).
Others: Co-organizer of Learning on Graph Conference (LoG 2025, 2024, 2023) local meetup, Cornell CS PhD Admission Committee Member (2025, 2024), Cornell CS Visit Day Czar.