Lingkai Kong
孔令恺
I am a postdoctoral fellow at Harvard, advised by Prof. Milind Tambe. I obtained my Ph.D. in Computational Science and Engineering from Georgia Institute of Technology, advised by Prof. Chao Zhang.
For more details, see my CV.
Many real-world decisions are high stakes, time sensitive, and made under uncertainty, with data that are limited, noisy, and high dimensional. My research advances AI methods for reliable and efficient decision making along three directions:
Diffusion models for representing complex, high-dimensional distributions, and large language models as priors that ground decisions in world knowledge.
Reinforcement learning for adaptive, long-horizon planning under uncertainty.
Deploying these methods in domains such as public health and sustainability.
Selected publications
- 2026Latent Spherical Flow Policy for Reinforcement Learning with Combinatorial ActionsIn International Conference on Machine Learning (ICML) (Spotlight), 2026
- 2026Reward Shaping for Inference-Time Alignment: A Stackelberg Game PerspectiveIn International Conference on Machine Learning (ICML), 2026
- 2025Composite Flow Matching for Reinforcement Learning with Shifted-Dynamics DataIn Advances in Neural Information Processing Systems (NeurIPS), 2025 (Spotlight)
- 2025Diffusion Models as Constrained Samplers for Optimization with Unknown ConstraintsIn Artificial Intelligence and Statistics (AISTATS), 2025
- 2024Aligning Large Language Models with Representation Editing: A Control PerspectiveIn Advances in Neural Information Processing Systems (NeurIPS), 2024
- 2023AdaPlanner: Adaptive Planning from Feedback with Language ModelsIn Advances in Neural Information Processing Systems (NeurIPS), 2023
- 2022End-to-End Stochastic Optimization with Energy-based ModelIn Advances in Neural Information Processing Systems (NeurIPS), 2022 (Oral, 181/10411, top 1.76%)