About

Xiaodong D Lin is a professor of Cognitive Science in Education at Teachers College, Columbia University, where she also serves as the Founding Director of the Education for Persistence and Innovation Center (EPIC). Her scholarly interests include education technology, science education, and the brain and cognition. Dr. Lin earned her Ph.D. in Instructional Psychology and Educational Technology from Purdue University, an M.A. in Elementary Science Education and Information Technology from Louisiana State University, and a B.A. in Foreign Language Education (English) and Linguistics from Teachers College, HeNan University. Her research focuses on motivating students to succeed despite obstacles in their academic and personal lives. A central theme in her work is understanding how learning about the struggles of scientists and other role models can enhance students' resilience and confidence in STEM learning. She incorporates stories of failure and persistence into curricula to improve students' ability to handle challenging learning tasks. Her work has been widely publicized, including by the American Psychological Association, NPR, CBS News, PBS, and international media. Dr. Lin has received numerous awards, such as the Career Achievement Award and Distinguished Research Award from the American Education Research Association, and was named a Carnegie Scholar. Recently, she was elected to serve on the education expert advisory board of the OECD and has contributed to the EDUCATION 2030 Initiative, advocating for the inclusion of self-enhancement in future education goals.

Research topics

• Artificial Intelligence
• Computer Science
• Natural Language Processing
• Information Retrieval
• Machine Learning
• Chemistry
• Biology
• Neuroscience
• Materials science
• Biophysics
• Speech recognition

Selected publications

• Hydrogel-Exosome Synergistic System: Mechanisms, Applications and Future Prospects in Regenerative Medicine

  Biomaterials Science · 2026-01-01

  article

  Hydrogels and exosomes are mutually beneficial and thus can be combined for regenerative medicine and innovative therapies. This strategy takes advantages of hydrogels to enhance the therapeutic potential of exosomes....

  PublisherDOI

• Impact of naturally occurring hemagglutinin substitutions on antigenicity and fitness of influenza A(H5N1) virus

  npj Viruses · 2025-10-01 · 3 citations

  articleOpen access

  In 2024, a human infection with clade 2.3.4.4b highly pathogenic avian influenza A(H5N1) virus was identified in the United States in an individual with no known exposure. Genetic analysis revealed two hemagglutinin (HA) substitutions, P136S and A156T, which may alter viral antigenicity. Virus isolation was unsuccessful, preventing timely serologic analysis. To overcome this limitation, we generated recombinant viruses by reverse genetics and characterized the effects of the substitutions on antigenicity, receptor binding, and replicative fitness. The A156T substitution introduced a potential N-linked glycosylation site, resulting in altered antigenicity and reduced replication in primary human nasal epithelial cells and ferrets. Importantly, the A(H5N1) candidate vaccine virus (CVV) IDCDC-RG80A, which possesses HA-T156, remained antigenically effective against viruses with and without these substitutions. These findings highlight the importance of sequencing, reverse-genetics approaches, and antigenically similar CVVs such as IDCDC-RG80A, for pandemic preparedness against evolving clade 2.3.4.4b A(H5N1) viruses.

  PublisherOA PDFDOI

• Patterns of shelf margin clinoform: control of the development of deep-water sedimentary systems

  Frontiers in Marine Science · 2025-02-24

  articleOpen accessSenior author

  The clinothem is the fundamental element of basin infill and plays an important role in the source-to-sink system of deep-water basins. In this study, which is based on 2D and 3D seismic data, the spatiotemporal evolution of clinothems and depositional systems from the Miocene to the Pleistocene in the Qiongdongnan Basin, northern South China Sea, is investigated. The following conclusions are drawn: (1) three variations of clinothems in the Qiongdongnan Basin were recognized respective O-, S- and F-type clinothems; (2) fluctuations in the relative base level, in combination with variations in sediment supply, result in different clinothem patterns that may be used to understand changes in these depositional factors; and (3) the development of O-type clinothem is usually accompanied by slope instability and slumping, leading to mass transport deposits on the basin plain. When suitable transport pathways exist (e.g., shelf canyons) on the shelf-edge or when the F-type clinothem develops, sediments can be dispersed basinward, promoting submarine fan development. Results from this study will help in understanding the sedimentological development of slope and basin plain areas and offers significant insights into the understanding of deep-sea depositional systems.

  PublisherOA PDFDOI

• Inter-view contrastive learning and miRNA fusion for lncRNA-protein interaction prediction in heterogeneous graphs

  Briefings in Bioinformatics · 2025-03-01 · 1 citations

  articleOpen access

  Predicting long non-coding RNA (lncRNA)-protein interactions is essential for understanding biological processes and discovering new therapeutic targets. In this study, we propose a novel model based on inter-view contrastive learning and miRNA fusion for lncRNA-protein interaction (LPI) prediction, called ICMF-LPI, which utilizes a heterogeneous information network to enhance LPI prediction. The model integrates miRNA as a mediator, constructing an lncRNA-miRNA-protein network, and employs metapath to extract diverse relationships from heterogeneous graphs. By fusing miRNA-related information and leveraging contrastive learning across inter-views, ICMF-LPI effectively captures potential interactions. Experimental results, including five-fold cross-validation, demonstrate the model's superior performance compared to several state-of-the-art methods, with significant improvements in the area under the receiver operating characteristic curve and the area under the precision-recall curve metrics. Notably, even when direct LPI connections are excluded, ICMF-LPI still achieves competitive predictive accuracy, performing comparably or better than some existing models. This demonstrates that the proposed model is effective in scenarios where direct interaction data are unavailable. This approach offers a promising direction for developing predictive models in bioinformatics, particularly in challenging conditions.

  PublisherOA PDFDOI

• Shallow gas accumulation mechanism in the Qiongdongnan Basin, South China Sea

  Marine and Petroleum Geology · 2025-03-19 · 6 citations

  articleOpen accessSenior author

  The Qiongdongnan Basin is one of the most important oil and gas basins in the South China Sea. Previous studies primarily focused on the middle strata around 3000–4000 m. In recent years, abundant shallow oil and gas resources have been discovered in the shallow Ledong Formation of the Qiongdongnan Basin. Previous researchers believed that these shallow accumulations originated from deep strata and that the shallow strata could not serve as effective source rocks. There was a lack of relevant studies on how deep gas sources transport gas to shallow areas, where gas was accumulated in shallow layers, how the cap rock was distributed, and how the transport system, reservoir, and cap rock matched. This research aimed to clarify the types and spatial distribution of transport systems in the study area using 3D seismic data , combined with methods such as root mean square amplitude attributes and variance cube slicing. Simultaneously, it sought to identify the development types and sizes of reservoirs and cap rocks and elucidated the relationships among these three influencing factors, clarifying the process of deep natural gas accumulation in shallow reservoirs. Six transport systems, including diapiric gas chimneys, fracture-type gas chimneys, polygonal faults and fractures, large faults, submarine slump, and multi-stage channels, were identified in this study, along with their distributions. The main reservoirs in the Ledong Formation were submarine fans , slope fans, and shoreland sandbars, with their upper portions predominantly capped by mass transport deposits and mudstone layers. Based on these findings, the spatial matching of transport systems and reservoirs was conducted to clarify the process of gas migration, transport, and accumulation from deep to shallow strata, summarizing corresponding accumulation models. Gas chimneys and faults were considered the primary transport systems, while submarine fans were identified as the highest-quality reservoirs, and mass transport deposits were effective cap rocks. The accumulation model formed by the combination of these three elements represented the primary accumulation mode in the study area. The results of this study will provide guidance for the development of shallow resources in the South China Sea. • Gas chimneys and faults were efficient transport systems. • Submarine fans were high-quality reservoir. • Mass transport deposits and mudstones were effective cap rocks. • The accumulation process of deep natural gas in shallow layers was clarified.

  PublisherDOI

• SDA-YOLO: An Object Detection Method for Peach Fruits in Complex Orchard Environments

  Sensors · 2025-07-17 · 6 citations

  articleOpen access1st authorCorresponding

  To address the challenges of leaf-branch occlusion, fruit mutual occlusion, complex background interference, and scale variations in peach detection within complex orchard environments, this study proposes an improved YOLOv11n-based peach detection method named SDA-YOLO. First, in the backbone network, the LSKA module is embedded into the SPPF module to construct an SPPF-LSKA fusion module, enhancing multi-scale feature representation for peach targets. Second, an MPDIoU-based bounding box regression loss function replaces CIoU to improve localization accuracy for overlapping and occluded peaches. The DyHead Block is integrated into the detection head to form a DMDetect module, strengthening feature discrimination for small and occluded targets in complex backgrounds. To address insufficient feature fusion flexibility caused by scale variations from occlusion and illumination differences in multi-scale peach detection, a novel Adaptive Multi-Scale Fusion Pyramid (AMFP) module is proposed to enhance the neck network, improving flexibility in processing complex features. Experimental results demonstrate that SDA-YOLO achieves precision (P), recall (R), mAP@0.95, and mAP@0.5:0.95 of 90.8%, 85.4%, 90%, and 62.7%, respectively, surpassing YOLOv11n by 2.7%, 4.8%, 2.7%, and 7.2%. This verifies the method's robustness in complex orchard environments and provides effective technical support for intelligent fruit harvesting and yield estimation.

  PublisherOA PDFDOI

• Platform value-added service strategy based on data factor endowment

  Electronic Commerce Research · 2025-04-21 · 1 citations

  article1st authorCorresponding
  PublisherDOI

• LOFT: Scalable and More Realistic Long-Context Evaluation

  2025-01-01

  articleOpen access

  Jinhyuk Lee, Anthony Chen, Zhuyun Dai, Dheeru Dua, Devendra Singh Sachan, Michael Boratko, Yi Luan, Séb Arnold, Vincent Perot, Siddharth Dalmia, Hexiang Hu, Xudong Lin, Panupong Pasupat, Aida Amini, Jeremy R. Cole, Sebastian Riedel, Iftekhar Naim, Ming-Wei Chang, Kelvin Guu. Findings of the Association for Computational Linguistics: NAACL 2025. 2025.

  PublisherOA PDFDOI

• EML-SlowFast: A behavior recognition model for lion-head goose

  Poultry Science · 2025-05-02 · 1 citations

  articleOpen accessSenior author

  The behavior of lion-head goose has a significant impact on their health status, activity levels, and productivity. It is therefore important to monitor the behavior of lion-head geese to enhance their health status, reproductive performance, and overall productivity. However, there is currently no specific behavioral recognition method for lion-head goose, which presents a significant challenge in quickly and effectively identifying various behaviors. To address this issue, this study proposes a model called EML-SlowFast, which is an improvement based on SlowFast. The model is capable of distinguishing five basic behaviors of lion-head goose: feeding, resting, preening, standing, and walking. The Efficient Channel Attention Bottleneck (ECAbneck) module and the Large Kernel Global-Local Feature Extraction (LGLE) module are designed and incorporated into the model. By combining and filtering channel information, the ECAbneck module enhances the model's ability to extract static characteristics from lion-head goose, increasing the accuracy of behavior recognition. The LGLE module captures temporal dependencies in lion-head goose behavior by integrating and extracting local and global features, thereby reinforcing the model's ability to model long-term temporal characteristics and further increasing accuracy. The experiment results showed that the average F1 score, average Precision, Accuracy, and average Recall of the EML-SlowFast model were 92.06 %, 91.60 %, 91.85 %, and 92.78 %, respectively, reflecting improvements of 4.03 %, 3.79 %, 4.14 %, and 4.45 % over the corresponding metrics of the SlowFast model. Furthermore, the FLOPs of the EML-SlowFast model was 10.807 G, which was a reduction of 7.358 G compared to the SlowFast model. Compared to commonly used behavior recognition models, the EML-SlowFast model has effective recognition of lion-head goose behaviors while maintaining low computational complexity, which is beneficial for deployment and use in scenarios with low computational resources. The EML-SlowFast model can rapidly and accurately recognize lion-head goose behaviors, providing a valuable reference for precision farming, reproduction, and health welfare monitoring of lion-head goose.

  PublisherDOI

• PuzzleGPT: Emulating Human Puzzle-Solving Ability for Time and Location Prediction

  ArXiv.org · 2025-01-24

  preprintOpen access

  The task of predicting time and location from images is challenging and requires complex human-like puzzle-solving ability over different clues. In this work, we formalize this ability into core skills and implement them using different modules in an expert pipeline called PuzzleGPT. PuzzleGPT consists of a perceiver to identify visual clues, a reasoner to deduce prediction candidates, a combiner to combinatorially combine information from different clues, a web retriever to get external knowledge if the task can't be solved locally, and a noise filter for robustness. This results in a zero-shot, interpretable, and robust approach that records state-of-the-art performance on two datasets -- TARA and WikiTilo. PuzzleGPT outperforms large VLMs such as BLIP-2, InstructBLIP, LLaVA, and even GPT-4V, as well as automatically generated reasoning pipelines like VisProg, by at least 32% and 38%, respectively. It even rivals or surpasses finetuned models.

  PublisherOA PDFDOI

Frequent coauthors

• Peng Shi

  City University of Hong Kong, Shenzhen Research Institute

  46 shared

• Shih‐Fu Chang

  45 shared

• Manling Li

  22 shared

• Yulei Niu

  15 shared

• Heng Ji

  14 shared

• Heng Ji

  13 shared

• Mike Zheng Shou

  National University of Singapore

  11 shared

• Guanyu Cai

  National University of Singapore

  11 shared

Education

• Ph.D., Instructional Psychology and Educational Technology

  Purdue University

• M.A., Elementary Science Education and Information Technology

  Louisiana State University

• B.A., Foreign Language Education (English) and Linguistics

  Teachers College, HeNan University

Awards & honors

• Career Achievement Award by the American Education Research…
• Distinguished Research Award by the American Education Resea…
• Carnegie Scholar by the Carnegie Corporation of New York