Denghong Luan (栾登闳)

About Me

I am an undergraduate student at the University of Science and Technology of China (USTC), School of the Gifted Young, majoring in Artificial Intelligence and Data Science. I am currently visiting the University of Hong Kong (HKU) as a Fung Scholar (Jan–May 2026), a fellowship awarded by the Victor and William Fung Foundation.

My research is primarily conducted at the Lab for Data Science(LDS), USTC, where I work on benchmarking and evaluating the reasoning capabilities of large language models. I am broadly interested in how LLMs reason about structured scientific problems and how to measure that reasoning faithfully.

Research Interests

• LLM Evaluation & Benchmarking  —  Building contamination-free, process-level benchmarks for assessing the scientific reasoning fidelity of frontier LLMs (PPRBench, PhysGrader)
• AI for Science  —  Applying knowledge-augmented LLMs to spectroscopy and other scientific identification tasks (Fingerprint Reasoner)
• Reinforcement Learning for LLMs  —  Exploring GRPO with process reward signals to improve step-by-step reasoning
• Indoor Localization  —  WiFi/CSI-based positioning via Channel State Information and simulated annealing optimization

Selected Publications

PPRBench: A Process-level Benchmark for LLMs’ Physical Reasoning Denghong Luan, Wentao Shi, Wenjie Wang, Xiangnan He Submitted to SIGIR 2026 (Resource Track), under review

An Integrated Simulated Annealing Algorithm for WiFi-based Indoor Positioning Denghong Luan, Yiheng Zhang, Chen Bin 30th International Computing and Combinatorics Conference (COCOON 2024)

News