Questions.
Methods. Discoveries.

@inproceedings{liu2026viper,
  abbr = {NeurIPS},
  title = {Mimicking the Physicist's Eye: A VLM-centric Approach for Physics Formula Discovery},
  author = {Liu, Jiaqi and Lai, Songning and Li, Pengze and Yu, Di and Zhou, Wenjie and Zhou, Yiyang and Xia, Peng and Wang, Zijun and Chen, Xi and Tang, Shixiang and Bai, Lei and Ouyang, Wanli and Ding, Mingyu and Yao, Huaxiu and Wang, Aoran},
  year = {2026},
  arxiv = {2508.17380},
  pdf = {https://arxiv.org/pdf/2508.17380},
  website = {https://jiaaqiliu.github.io/VIPER-R1/},
  selected = {true},
  booktitle = {Advances in Neural Information Processing Systems},
  note = {Accepted}
}

@article{li2026coupled_harness_model,
  abbr = {arXiv},
  title = {Coupled Harness--Model Evolution for Scientific Equation Discovery},
  author = {Li, Pengze and Zhang, Jiaquan and Su, Encheng and Long, Yunbo and Zeng, Zihang and Liu, Xinping and Zhang, Han and Yu, Junchi and Tang, Shixiang and Wang, Aoran and Chen, Xi},
  journal = {arXiv preprint arXiv:2610.17740},
  year = {2026},
  arxiv = {2610.17740},
  pdf = {https://arxiv.org/pdf/2610.17740},
  website = {https://arxiv.org/abs/2610.17740}
}

@article{su2026protean,
  abbr = {arXiv},
  title = {PROTEAN: Beyond De Novo Generation - Can LLM Agents Perform Text-Guided Protein Structure Editing?},
  author = {Su, Encheng and Liu, Xinping and Wang, Yizhou and Zhong, Bozitao and Li, Pengze and Wu, Jianyu and Li, Mingxiao and Ma, Xinzhu and Wang, Aoran and Hong, Liang and Ouyang, Wanli and Tang, Chen and Li, Houqiang and Tang, Shixiang},
  journal = {arXiv preprint arXiv:2610.17765},
  year = {2026},
  arxiv = {2610.17765},
  pdf = {https://arxiv.org/pdf/2610.17765},
  website = {https://arxiv.org/abs/2610.17765}
}

@article{su2026bridge,
  abbr = {arXiv},
  title = {BRIDGE: Selective Boundary Repair with Dependency Graph Evidence in Multi-Agent Systems},
  author = {Su, Encheng and Li, Pengze and Hua, Ziheng and Yu, Qiucheng and Wu, Jianyu and Tang, Chen and Wang, Yizhou and Wang, Lintao and Xiang, Junjia and Ma, Xinzhu and Chen, Xi and Tang, Shixiang and Li, Houqiang and Chen, Xi and Wang, Aoran},
  journal = {arXiv preprint arXiv:2610.17805},
  year = {2026},
  arxiv = {2610.17805},
  pdf = {https://arxiv.org/pdf/2610.17805},
  website = {https://arxiv.org/abs/2610.17805}
}

@article{tang2026native_structural_reasoning,
  abbr = {arXiv},
  title = {Accurate, Interdisciplinary and Transparent Structure-property Understanding with Deep Native Structural Reasoning},
  author = {Tang, Chen and Wang, Yizhou and Wu, Jianyu and Wang, Lintao and Tang, Shixiang and Li, Pengze and Su, Encheng and Yao, Jun and Xiao, Jiabei and Shi, Yuqi and Li, Jielan and Hao, Hongxia and Gao, Zhangyang and Wu, Fang and Fei, Ben and Yue, Xiangyu and Tan, Pan and Zhong, Bozitao and Zhang, Jinouwen and Wang, Aoran and Lu, Yan and Liu, Jiaheng and Ma, Xinzhu and Hong, Liang and Zheng, Mingyue and Torr, Phil and Zhou, Bowen and Ouyang, Wanli and Bai, Lei},
  journal = {arXiv preprint arXiv:2607.07708},
  year = {2026},
  arxiv = {2607.07708},
  pdf = {https://arxiv.org/pdf/2607.07708},
  website = {https://arxiv.org/abs/2607.07708},
  selected = {true}
}

@inproceedings{su2026s3mem,
  abbr = {EMNLP},
  title = {S3Mem: Structured Spatiotemporal Scene-Event Memory for Long-Horizon Interactive Question Answering},
  author = {Su, Encheng and Wu, Jianyu and Zhang, Jinouwen and Yu, Qiucheng and Tang, Chen and Li, Pengze and Wang, Lintao and Wang, Aoran and Ma, Xinzhu and Tang, Shixiang and Wang, Yizhou and Li, Houqiang},
  booktitle = {Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (Main Conference)},
  note = {Accepted},
  year = {2026},
  arxiv = {2605.28831},
  pdf = {https://arxiv.org/pdf/2605.28831},
  website = {https://arxiv.org/abs/2605.28831}
}

@article{wang2026sigfn,
  abbr = {Mach. Learn.},
  title = {Harnessing Generative Flow Networks for Effective Structural Inference},
  author = {Wang, Aoran and Pang, Jun},
  journal = {Machine Learning},
  year = {2026},
  volume = {115},
  number = {5},
  pages = {113},
  doi = {10.1007/s10994-025-06985-w},
  website = {https://link.springer.com/article/10.1007/s10994-025-06985-w}
}

@article{li2026hidden_plain_sight,
  abbr = {arXiv},
  title = {Hidden in Plain Sight: Visual-to-Symbolic Analytical Solution Inference from Field Visualizations},
  author = {Li, Pengze and Zhang, Jiaquan and Long, Yunbo and Liu, Xinping and Zhou, Wenjie and Su, Encheng and Zeng, Zihang and Liu, Jiaqi and Liu, Jiyao and Yu, Junchi and Liu, Lihao and Torr, Philip and Tang, Shixiang and Wang, Aoran and Chen, Xi},
  journal = {arXiv preprint arXiv:2604.08863},
  year = {2026},
  arxiv = {2604.08863},
  pdf = {https://arxiv.org/pdf/2604.08863},
  website = {https://arxiv.org/abs/2604.08863}
}

@inproceedings{wan2026deepresearch_arena,
  abbr = {AAAI},
  title = {Deep Research Arena: The First Exam of LLMs' Research Abilities via Seminar-Grounded Tasks},
  author = {Wan, Haiyuan and Yang, Chen and Yu, Junchi and Tu, Meiqi and Lu, Jiaxuan and Yu, Di and Cao, Jianbao and Gao, Ben and Xie, Jiaqing and Wang, Aoran and Zhang, Wenlong and Torr, Philip and Zhou, Dongzhan},
  year = {2026},
  arxiv = {2509.01396},
  pdf = {https://arxiv.org/pdf/2509.01396},
  website = {https://ojs.aaai.org/index.php/AAAI/article/view/40620},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  volume = {40},
  number = {39},
  pages = {33341--33349},
  doi = {10.1609/aaai.v40i39.40620}
}

@article{su2026sciif,
  abbr = {arXiv},
  title = {SciIF: Benchmarking Scientific Instruction Following Towards Rigorous Scientific Intelligence},
  author = {Su, Encheng and Wu, Jianyu and Tang, Chen and Wang, Lintao and Li, Pengze and Wang, Aoran and Zhang, Jinouwen and Wang, Yizhou and Meng, Yuan and Ma, Xinzhu and Tang, Shixiang and Li, Houqiang},
  journal = {arXiv preprint arXiv:2601.04770},
  year = {2026},
  arxiv = {2601.04770},
  pdf = {https://arxiv.org/pdf/2601.04770},
  website = {https://arxiv.org/abs/2601.04770}
}

@inproceedings{tong2025kolmogorovarnoldnetworkgeneregulatory,
  abbr = {CMSB},
  title = {Kolmogorov-Arnold Network for Gene Regulatory Network Inference},
  author = {Tong, Tsz Pan and Wang, Aoran and Panagopoulos, George and Pang, Jun},
  booktitle = {Computational Methods in Systems Biology (CMSB 2025)},
  year = {2026},
  volume = {15959},
  pages = {221--239},
  series = {Lecture Notes in Computer Science},
  publisher = {Springer},
  doi = {10.1007/978-3-032-01436-8_12},
  website = {https://link.springer.com/chapter/10.1007/978-3-032-01436-8_12},
  pdf = {https://arxiv.org/pdf/2506.13740},
  arxiv = {2506.13740},
  eprint = {2506.13740},
  archiveprefix = {arXiv},
  primaryclass = {cs.CE},
  url = {https://arxiv.org/abs/2506.13740}
}

@inproceedings{tong2024integrating,
  abbr = {ECML PKDD},
  title = {Integrating Optimal Transport and Structural Inference Models for GRN Inference from Single-cell Data},
  author = {Tong, Tsz Pan and Wang, Aoran and Panagopoulos, George and Pang, Jun},
  booktitle = {Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML PKDD 2024 Workshops)},
  year = {2026},
  volume = {2560},
  pages = {257--271},
  series = {Communications in Computer and Information Science},
  publisher = {Springer},
  doi = {10.1007/978-3-032-25311-8_21},
  website = {https://link.springer.com/chapter/10.1007/978-3-032-25311-8_21},
  arxiv = {2409.15080},
  pdf = {https://arxiv.org/pdf/2409.15080},
  abstract = {We introduce a novel gene regulatory network (GRN) inference method that integrates optimal transport (OT) with a deep-learning structural inference model. Advances in next-generation sequencing enable detailed yet destructive gene expression assays at the single-cell level, resulting in the loss of cell evolutionary trajectories. Due to technological and cost constraints, single-cell experiments often feature cells sampled at irregular and sparse time points with a small sample size. Although trajectory-based structural inference models can accurately reveal the underlying interaction graph from observed data, their efficacy depends on the inputs of thousands of regularly sampled trajectories. The irregularly-sampled nature of single-cell data precludes the direct use of these powerful models for reconstructing GRNs. Optimal transport, a classical mathematical framework that minimize transportation costs between distributions, has shown promise in multi-omics data integration and cell fate prediction. Utilizing OT, our method constructs mappings between consecutively sampled cells to form cell-level trajectories, which are given as input to a structural inference model that recovers the GRN from single-cell data. Through case studies in two synthetic datasets, we demonstrate the feasibility of our proposed method and its promising performance over eight state-of-the-art GRN inference methods.}
}

@inproceedings{dai2025sequence_structure,
  abbr = {NeurIPS},
  title = {From Sequence to Structure: Uncovering Substructure Reasoning in Transformers},
  author = {Dai, Xinnan and Yang, Kai and Revolinsky, Jay and Guo, Kai and Wang, Aoran and Zhang, Bohang and Tang, Jiliang},
  booktitle = {Advances in Neural Information Processing Systems},
  year = {2025},
  volume = {38},
  doi = {10.52202/085713-2552},
  pdf = {https://proceedings.neurips.cc/paper_files/paper/2025/file/6def705c0557104daf581c19b9e2e5e3-Paper-Conference.pdf},
  website = {https://proceedings.neurips.cc/paper_files/paper/2025/hash/6def705c0557104daf581c19b9e2e5e3-Abstract-Conference.html}
}

@article{hu2025scientific_llm_survey,
  abbr = {arXiv},
  title = {A Survey of Scientific Large Language Models: From Data Foundations to Agent Frontiers},
  author = {Hu, Ming and Ma, Chenglong and Li, Wei and Xu, Wanghan and Wu, Jiamin and Hu, Jucheng and Li, Tianbin and Zhuang, Guohang and Liu, Jiaqi and Lu, Yingzhou and Chen, Ying and Zhang, Chaoyang and Tan, Cheng and Ying, Jie and Wu, Guocheng and Gao, Shujian and Chen, Pengcheng and Lin, Jiashi and Wu, Haitao and Chen, Lulu and Wang, Fengxiang and Zhang, Yuanyuan and Zhao, Xiangyu and Tang, Feilong and Su, Encheng and Ning, Junzhi and Liu, Xinyao and Du, Ye and Ji, Changkai and Jiang, Pengfei and Tang, Cheng and Huang, Ziyan and Liu, Jiyao and Wei, Jiaqi and Yang, Yuejin and Zhang, Xiang and Wang, Guangshuai and Yang, Yue and Xu, Huihui and Chen, Ziyang and Wang, Yizhou and Tang, Chen and Wu, Jianyu and Ren, Yuchen and Yan, Siyuan and Wang, Zhonghua and Xu, Zhongxing and Su, Shiyan and Sun, Shangquan and Zhao, Runkai and Zhang, Zhisheng and Yang, Dingkang and Wei, Jinjie and Wang, Jiaqi and Xu, Jiahao and Yan, Jiangtao and Tang, Wenhao and Zhu, Hongze and Liu, Yu and Wang, Fudi and Shen, Yiqing and Ji, Yuanfeng and Su, Yanzhou and Xie, Tong and Shan, Hongming and Feng, Chun-Mei and Hou, Zhi and Song, Diping and Liu, Lihao and Huang, Yanyan and Yu, Lequan and Fu, Bin and Wang, Shujun and Li, Xiaomeng and Hu, Xiaowei and Gu, Yun and Fei, Ben and Wang, Benyou and Cao, Yuewen and Shen, Minjie and Xu, Jie and Duan, Haodong and Yan, Fang and Hao, Hongxia and Li, Jielan and Du, Jiajun and Wang, Yanbo and Razzak, Imran and Deng, Zhongying and Zhang, Chi and Wu, Lijun and He, Conghui and Lu, Zhaohui and Huang, Jinhai and Shao, Wenqi and Liu, Yihao and Luo, Siqi and Xin, Yi and Liu, Xiaohong and Ling, Fenghua and Li, Yuqiang and Wang, Aoran and Sun, Siqi and Zheng, Qihao and Dong, Nanqing and Fu, Tianfan and Zhou, Dongzhan and Lu, Yan and Zhang, Wenlong and Ye, Jin and Cai, Jianfei and Chen, Yirong and Ouyang, Wanli and Qiao, Yu and Ge, Zongyuan and Tang, Shixiang and He, Junjun and Song, Chunfeng and Bai, Lei and Zhou, Bowen},
  journal = {arXiv preprint arXiv:2508.21148},
  year = {2025},
  arxiv = {2508.21148},
  pdf = {https://arxiv.org/pdf/2508.21148},
  website = {https://arxiv.org/abs/2508.21148}
}

@article{bai2025interns1,
  abbr = {arXiv},
  title = {Intern-S1: A Scientific Multimodal Foundation Model},
  author = {Bai, Lei and Cai, Zhongrui and Cao, Yuhang and Cao, Maosong and Cao, Weihan and Chen, Chiyu and Chen, Haojiong and Chen, Kai and Chen, Pengcheng and Chen, Ying and Chen, Yongkang and Cheng, Yu and Chu, Pei and Chu, Tao and Cui, Erfei and Cui, Ganqu and Cui, Long and Cui, Ziyun and Deng, Nianchen and Ding, Ning and Dong, Nanqing and Dong, Peijie and Dou, Shihan and Du, Sinan and Duan, Haodong and Fan, Caihua and Gao, Ben and Gao, Changjiang and Gao, Jianfei and Gao, Songyang and Gao, Yang and Gao, Zhangwei and Ge, Jiaye and Ge, Qiming and Gu, Lixin and Gu, Yuzhe and Guo, Aijia and Guo, Qipeng and Guo, Xu and He, Conghui and He, Junjun and Hong, Yili and Hou, Siyuan and Hu, Caiyu and Hu, Hanglei and Hu, Jucheng and Hu, Ming and Hua, Zhouqi and Huang, Haian and Huang, Junhao and Huang, Xu and Huang, Zixian and Jiang, Zhe and Kong, Lingkai and Li, Linyang and Li, Peiji and Li, Pengze and Li, Shuaibin and Li, Tianbin and Li, Wei and Li, Yuqiang and Lin, Dahua and Lin, Junyao and Lin, Tianyi and Lin, Zhishan and Liu, Hongwei and Liu, Jiangning and Liu, Jiyao and Liu, Junnan and Liu, Kai and Liu, Kaiwen and Liu, Kuikun and Liu, Shichun and Liu, Shudong and Liu, Wei and Liu, Xinyao and Liu, Yuhong and Liu, Zhan and Lu, Yinquan and Lv, Haijun and Lv, Hongxia and Lv, Huijie and Lv, Qitan and Lv, Ying and Lyu, Chengqi and Ma, Chenglong and Ma, Jianpeng and Ma, Ren and Ma, Runmin and Ma, Runyuan and Ma, Xinzhu and Ma, Yichuan and Ma, Zihan and Mi, Sixuan and Ning, Junzhi and Ning, Wenchang and Pang, Xinle and Peng, Jiahui and Peng, Runyu and Qiao, Yu and Qiu, Jiantao and Qu, Xiaoye and Qu, Yuan and Ren, Yuchen and Shang, Fukai and Shao, Wenqi and Shen, Junhao and Shen, Shuaike and Song, Chunfeng and Song, Demin and Song, Diping and Su, Chenlin and Su, Weijie and Sun, Weigao and Sun, Yu and Tan, Qian and Tang, Cheng and Tang, Huanze and Tang, Kexian and Tang, Shixiang and Tong, Jian and Wang, Aoran and Wang, Bin and Wang, Dong and Wang, Lintao and Wang, Rui and Wang, Weiyun and Wang, Wenhai and Wang, Jiaqi and Wang, Yi and Wang, Ziyi and Wu, Ling-I and Wu, Wen and Wu, Yue and Wu, Zijian and Xiao, Linchen and Xing, Shuhao and Xu, Chao and Xu, Huihui and Xu, Jun and Xu, Ruiliang and Xu, Wanghan and Yang, GanLin and Yang, Yuming and Ye, Haochen and Ye, Jin and Ye, Shenglong and Yu, Jia and Yu, Jiashuo and Yu, Jing and Yuan, Fei and Zang, Yuhang and Zhang, Bo and Zhang, Chao and Zhang, Chen and Zhang, Hongjie and Zhang, Jin and Zhang, Qiaosheng and Zhang, Qiuyinzhe and Zhang, Songyang and Zhang, Taolin and Zhang, Wenlong and Zhang, Wenwei and Zhang, Yechen and Zhang, Ziyang and Zhao, Haiteng and Zhao, Qian and Zhao, Xiangyu and Zhao, Xiangyu and Zhou, Bowen and Zhou, Dongzhan and Zhou, Peiheng and Zhou, Yuhao and Zhou, Yunhua and Zhu, Dongsheng and Zhu, Lin and Zou, Yicheng},
  journal = {arXiv preprint arXiv:2508.15763},
  year = {2025},
  arxiv = {2508.15763},
  pdf = {https://arxiv.org/pdf/2508.15763},
  website = {https://arxiv.org/abs/2508.15763},
  selected = {true}
}

@article{wang2025physunibench,
  abbr = {arXiv},
  title = {PhysUniBench: A Multi-Modal Physics Reasoning Benchmark at Undergraduate Level},
  author = {Wang, Lintao and Su, Encheng and Liu, Jiaqi and Li, Pengze and Xiao, Jiabei and Zhang, Wenlong and Dai, Xinnan and Chen, Xi and Meng, Yuan and Bai, Lei and Ouyang, Wanli and Tang, Shixiang and Wang, Aoran and Ma, Xinzhu},
  journal = {arXiv preprint arXiv:2506.17667},
  year = {2025},
  arxiv = {2506.17667},
  pdf = {https://arxiv.org/pdf/2506.17667},
  website = {https://prismax-team.github.io/PhysUniBenchmark/},
  selected = {true}
}

@article{liang2025mc4ad,
  abbr = {arXiv},
  title = {Examining the Source of Defects from a Mechanical Perspective for 3D Anomaly Detection},
  author = {Liang, Hanzhe and Wang, Aoran and Zhou, Jie and Jin, Xin and Gao, Can and Wang, Jinbao},
  journal = {arXiv preprint arXiv:2505.05901},
  year = {2025},
  arxiv = {2505.05901},
  pdf = {https://arxiv.org/pdf/2505.05901},
  website = {https://arxiv.org/abs/2505.05901}
}

@inproceedings{wang-2025-sgsi,
  abbr = {ICML},
  title = {Guided Structural Inference: Leveraging Priors with Soft Gating Mechanisms},
  author = {Wang, Aoran and Dai, Xinnan and Pang, Jun},
  booktitle = {Proceedings of the 42nd International Conference on Machine Learning},
  year = {2025},
  volume = {267},
  pages = {62761--62783},
  series = {Proceedings of Machine Learning Research},
  publisher = {PMLR},
  website = {https://proceedings.mlr.press/v267/wang25w.html},
  pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/wang25w/wang25w.pdf},
  abstract = {Existing methods for inferring latent relational structures struggle to integrate partial prior knowledge, such as known edges, node-degree constraints, and global sparsity, without destabilizing training or conflicting with probabilistic assumptions. We propose Soft-Gated Structural Inference (SGSI), a VAE framework that seamlessly incorporates domain constraints via (1) soft gating with learnable edge masks to preserve gradients, (2) cloning-clamping of deterministic edges to avoid distributional conflicts, and (3) adaptive regularization to balance data-driven learning with domain constraints. By excluding known edges from stochastic inference, SGSI reallocates capacity to uncertain interactions, optimizing the information bottleneck trade-off. Experiments on 16 datasets show SGSI improves edge recovery by up to 9% AUROC over baselines, scales to larger graphs (94.2% AUROC), and maintains stable training. SGSI bridges domain expertise with data-driven learning, enabling interpretable and robust structural discovery in dynamical systems.}
}

@inproceedings{wang2024sicsm,
  abbr = {NeurIPS},
  title = {Structural Inference of Dynamical Systems with Conjoined State Space Models},
  author = {Wang, Aoran and Pang, Jun},
  booktitle = {The Thirty-eighth Annual Conference on Neural Information Processing Systems},
  year = {2024},
  website = {https://proceedings.neurips.cc/paper_files/paper/2024/hash/89c61fce5a8b73871d1c4073f486b134-Abstract-Conference.html},
  pdf = {https://proceedings.neurips.cc/paper_files/paper/2024/file/89c61fce5a8b73871d1c4073f486b134-Paper-Conference.pdf},
  abstract = {This paper introduces SICSM, a novel structural inference framework that integrates Selective State Space Models (selective SSMs) with Generative Flow Networks (GFNs) to handle the challenges posed by dynamical systems with irregularly sampled trajectories and partial observations. By utilizing the robust temporal modeling capabilities of selective SSMs, our approach learns input-dependent transition functions that adapt to non-uniform time intervals, thereby enhancing the accuracy of structural inference. By aggregating dynamics across diverse temporal dependencies and channeling them into the GFN, the SICSM adeptly approximates the posterior distribution of the system's structure. This process not only enables precise inference of complex interactions within partially observed systems but also ensures the seamless integration of prior knowledge, enhancing the model’s accuracy and robustness. Extensive evaluations on sixteen diverse datasets demonstrate that SICSM outperforms existing methods, particularly in scenarios characterized by irregular sampling and incomplete observations, which highlight its potential as a reliable tool for scientific discovery and system diagnostics in disciplines that demand precise modeling of complex interactions.}
}

@inproceedings{wang2024benchmarking,
  abbr = {NeurIPS},
  title = {Benchmarking Structural Inference Methods for Interacting Dynamical Systems with Synthetic Data},
  author = {Wang, Aoran and Tong, Tsz Pan and Mizera, Andrzej and Pang, Jun},
  booktitle = {The Thirty-eighth Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
  year = {2024},
  website = {https://proceedings.neurips.cc/paper_files/paper/2024/hash/f3c812da38d1bc796cb2e8235eee96bf-Abstract-Datasets_and_Benchmarks_Track.html},
  pdf = {https://proceedings.neurips.cc/paper_files/paper/2024/file/f3c812da38d1bc796cb2e8235eee96bf-Paper-Datasets_and_Benchmarks_Track.pdf},
  abstract = {Understanding complex dynamical systems begins with identifying their topological structures, which expose the organization of the systems. This requires robust structural inference methods that can deduce structure from observed behavior. However, existing methods are often domain-specific and lack a standardized, objective comparison framework. We address this gap by benchmarking 13 structural inference methods from various disciplines on simulations representing two types of dynamics and 11 interaction graph models, supplemented by a biological experimental dataset to mirror real-world application. We evaluated the methods for accuracy, scalability, robustness, and sensitivity to graph properties. Our findings indicate that deep learning methods excel with multi-dimensional data, while classical statistics and information theory based approaches are notably accurate and robust. Additionally, performance correlates positively with the graph's average shortest path length. This benchmark should aid researchers in selecting suitable methods for their specific needs and stimulate further methodological innovation.}
}

@inproceedings{wang-2024-sidec,
  abbr = {ICLR},
  title = {Structural Inference with Dynamics Encoding and Partial Correlation Coefficients},
  author = {Wang, Aoran and Pang, Jun},
  booktitle = {The Twelfth International Conference on Learning Representations},
  year = {2024},
  pdf = {https://proceedings.iclr.cc/paper_files/paper/2024/file/5be3783ea9d43d7add5409c101d87d83-Paper-Conference.pdf},
  website = {https://proceedings.iclr.cc/paper_files/paper/2024/hash/5be3783ea9d43d7add5409c101d87d83-Abstract-Conference.html},
  abstract = {This paper introduces a novel approach to structural inference, combining a variational dynamics encoder with partial correlation coefficients. In contrast to prior methods, our approach leverages variational inference to encode node dynamics within latent variables, and structural reconstruction relies on the calculation of partial correlation coefficients derived from these latent variables. This unique design endows our method with scalability and extends its applicability to both one-dimensional and multi-dimensional feature spaces. Furthermore, by reorganizing latent variables according to temporal steps, our approach can effectively reconstruct directed graph structures. We validate our method through extensive experimentation on twenty datasets from a benchmark dataset and biological networks. Our results showcase the superior scalability, accuracy, and versatility of our proposed approach compared to existing methods. Moreover, experiments conducted on noisy data affirm the robustness of our method.}
}

@inproceedings{pmlr-v202-wang23ak,
  abbr = {ICML},
  title = {Effective and Efficient Structural Inference with Reservoir Computing},
  author = {Wang, Aoran and Tong, Tsz Pan and Pang, Jun},
  booktitle = {Proceedings of the 40th International Conference on Machine Learning},
  pages = {36391--36410},
  year = {2023},
  volume = {202},
  series = {Proceedings of Machine Learning Research},
  publisher = {PMLR},
  pdf = {https://proceedings.mlr.press/v202/wang23ak/wang23ak.pdf},
  website = {https://proceedings.mlr.press/v202/wang23ak.html},
  abstract = {In this paper, we present an effective and efficient structural inference approach by integrating a Reservoir Computing (RC) network into a Variational Auto-encoder-based (VAE-based) structural inference framework. With the help of Bi-level Optimization, the backbone VAE-based method follows the Information Bottleneck principle and infers a general adjacency matrix in its latent space; the RC net substitutes the partial role of the decoder and encourages the whole approach to perform further steps of gradient descent based on limited available data. The experimental results on various datasets including biological networks, simulated fMRI data, and physical simulations show the effectiveness and efficiency of our proposed method for structural inference, either with much fewer trajectories or with much shorter trajectories compared with previous works.}
}

@inproceedings{pmlr-v202-wang23ac,
  abbr = {ICML},
  title = {Active Learning based Structural Inference},
  author = {Wang, Aoran and Pang, Jun},
  booktitle = {Proceedings of the 40th International Conference on Machine Learning},
  pages = {36224--36245},
  year = {2023},
  volume = {202},
  series = {Proceedings of Machine Learning Research},
  publisher = {PMLR},
  pdf = {https://proceedings.mlr.press/v202/wang23ac/wang23ac.pdf},
  website = {https://proceedings.mlr.press/v202/wang23ac.html},
  abstract = {In this paper, we propose a novel framework, Active Learning based Structural Inference (ALaSI), to infer the existence of directed connections from observed agents’ states over a time period in a dynamical system. With the help of deep active learning, ALaSI is competent in learning the representation of connections with a relatively small pool of prior knowledge. Moreover, based on information theory, the proposed inter- and out-of-scope message learning pipelines are remarkably beneficial to structural inference for large dynamical systems. We evaluate ALaSI on various large datasets including simulated systems and real-world networks, to demonstrate that ALaSI is able to outperform previous methods in precisely inferring the existence of connections in large systems under either supervised learning or unsupervised learning.}
}

@inproceedings{aoran2022isidg,
  abbr = {NeurIPS},
  author = {Wang, Aoran and Pang, Jun},
  booktitle = {Advances in Neural Information Processing Systems},
  pages = {8717--8730},
  title = {Iterative Structural Inference of Directed Graphs},
  website = {https://papers.nips.cc/paper_files/paper/2022/hash/39717429762da92201a750dd03386920-Abstract-Conference.html},
  pdf = {https://proceedings.neurips.cc/paper_files/paper/2022/file/39717429762da92201a750dd03386920-Paper-Conference.pdf},
  volume = {35},
  year = {2022},
  abstract = {In this paper, we propose a variational model, iterative Structural Inference of Directed Graphs (iSIDG), to infer the existence of directed interactions from observational agents’ features over a time period in a dynamical system. First, the iterative process in our model feeds the learned interactions back to encourage our model to eliminate indirect interactions and to emphasize directional representation during learning. Second, we show that extra regularization terms in the objective function for smoothness, connectiveness, and sparsity prompt our model to infer a more realistic structure and to further eliminate indirect interactions. We evaluate iSIDG on various datasets including biological networks, simulated fMRI data, and physical simulations to demonstrate that our model is able to precisely infer the existence of interactions, and is significantly superior to baseline models.}
}

@inproceedings{9294630,
  abbr = {ITSC},
  author = {Hu, Haohao and Wang, Aoran and Sons, Marc and Lauer, Martin},
  booktitle = {2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC)},
  title = {ViPNet: An End-to-End 6D Visual Camera Pose Regression Network},
  year = {2020},
  volume = {},
  number = {},
  pages = {1-7},
  doi = {10.1109/ITSC45102.2020.9294630},
  website = {https://ieeexplore.ieee.org/document/9294630},
  abstract = {In this work, we present a visual pose regression network: ViPNet. It is robust and real-time capable on mobile platforms such as self-driving vehicles. We train a convolutional neural network to estimate the six degrees of freedom camera pose from a single monocular image in an end-to-end manner. In order to estimate camera poses with uncertainty, we use a Bayesian version of the ResNet-50 as our basic network. SEBlocks are applied in residual units to increase our model's sensitivity to informative features. Our ViPNet is trained using a geometric loss function with trainable parameters, which can simplify the fine-tuning process significantly. We evaluate our ViPNet on the Cambridge Landmarks dataset and also on our Karl-Wilhelm-Plaza dataset, which is recorded with an experimental vehicle. As evaluation results, our ViPNet outperforms other end-to-end monocular camera pose estimation methods. Our ViPNet requires only 9-15ms to predict one camera pose, which allows us to run it with a very high frequency.}
}