Published in ICML 2025, 2025
A theoretical and learning-based end-to-end framework (FOLOC) for identification and optimal control of fractional-order LTI systems, with sample complexity guarantees and robust performance under non-Gaussian noise.
Published in IEEE LAD 2025, 2025
HDLCoRe is a training-free framework leveraging HDL-aware CoT prompting and a two-stage RAG system to mitigate hallucinations in LLM-generated HDL, significantly improving syntactic and functional correctness on RTLLM2.0.
Published in NeurIPS 2024, 2025
A novel structure-aware framework that integrates graph coarsening, node representation learning, and policy optimization for device placement on computation graphs, achieving up to ~60% inference speedup.