A research paper led by Shengjian Wu, Special Technical Adviser to the SINGAPORE-LAW-PRESS Research Institute, has been accepted to the Main Track of NeurIPS 2026, one of the world's leading international conferences in artificial intelligence. Wu is the first author of “Order Matters: Competition-Guided Query Ordering for RNN-Based Object Detection.”
Illustration: Official NeurIPS 2026 acceptance notification for the paper.
01 | NeurIPS: A Leading International Conference in Artificial Intelligence
Unlike disciplines in which journals are the principal medium of scholarly publication, important frontier research in artificial intelligence is often first presented at leading international conferences. Such conferences are not simply forums for academic exchange: rigorous peer review is used to select and formally publish original research.
NeurIPS (Conference on Neural Information Processing Systems) is among the most influential conferences in artificial intelligence worldwide. It is classified as a Category A international conference in artificial intelligence by the China Computer Federation (CCF), and received an A* rating—the highest classification—in the ICORE 2026 computer science conference rankings.
Many landmark research contributions in the development of artificial intelligence have appeared at NeurIPS.
For example, “Attention Is All You Need,” the influential 2017 NeurIPS paper, introduced the Transformer architecture now widely used in large models such as ChatGPT and DeepSeek.
That architecture has since been applied extensively to large language models, computer vision and multimodal AI, becoming a key foundation of modern artificial intelligence.
As a major publication venue for frontier AI research, NeurIPS attracts submissions every year from leading universities, research institutions and technology companies across the world.
In 2025, for example, the Main Track received more than 21,000 valid submissions, with a final acceptance rate of approximately 24.5%.
Research from institutions including Google, Tsinghua University, Peking University, Stanford University, the Massachusetts Institute of Technology, Meta and Microsoft has been published at the conference.
By bringing together leading AI researchers from around the world in academic exchange and competition, NeurIPS occupies an important and influential position in the global AI research community.
02 | First-Author Paper by Shengjian Wu Accepted to the NeurIPS 2026 Main Track
The accepted paper addresses object detection, a central problem in computer vision.
To tackle issues arising when multiple queries produce duplicate predictions for the same object and undermine training stability, the researchers propose DETRNN, a new method that introduces competition-guided query ordering into a recurrent refinement process.
By explicitly constructing a query order based on confidence and similarity, the method reduces redundant predictions, improves training stability and increases object detection accuracy.
Experiments across several DETR-style detectors demonstrate relatively consistent performance improvements.
The acceptance of Wu's first-author paper to the NeurIPS 2026 Main Track reflects his expertise and innovative research at the forefront of AI, as well as recognition of the work through the international academic review process.
03 | Connecting Frontier Technology with Legal Research and Professional Publishing
As a Special Technical Adviser to the SINGAPORE-LAW-PRESS Research Institute, Wu follows developments in AI and their intersections with legal scholarship and practice, contributing a technical perspective to the Institute's academic exchange and professional research.
The Research Institute values interdisciplinary collaboration between law and frontier technologies such as AI. It seeks to connect specialists from different disciplines and broaden the scope of legal technology research.
SINGAPORE-LAW-PRESS will continue to engage with developments at the intersection of technological innovation and legal practice, supporting knowledge dissemination, academic exchange and professional publishing.
Appendix | Abstract of the Paper (Original English)
DETR-style detectors use one-to-one bipartite matching during training to assign object queries to ground-truth objects, enabling end-to-end set prediction without non-maximum suppression (NMS). However, without an explicit de-duplication procedure, multiple queries can still produce highly similar hypotheses for the same object, making training unstable and predictions less decisive.
Inspired by the sequential ordering of NMS, we propose DETRNN, a plug-and-play module that turns unordered object queries into a competition-aware sequence for recurrent refinement.
DETRNN builds an explicit confidence-and-similarity based order from prior predictions, then refines queries with an RNN along this order to model competition inside the decoder.
This ordered recurrent refinement reduces redundant predictions, stabilizes optimization, and improves final detection accuracy. Experiments on multiple DETR-style detectors show consistent gains with comparable efficiency.
References: CCF conference rankings; ICORE 2026 Conference Portal; NeurIPS 2025 Main Track acceptance statistics.
Profile | Shengjian Wu
Special Technical Adviser, SINGAPORE-LAW-PRESS Research Institute
Shengjian Wu holds a bachelor's degree in software engineering and a master's degree in electronic information from East China Normal University. He has over a decade of experience in AI, computer vision, image processing, network data analysis and software systems development, including the implementation of technology projects. His areas of expertise include object detection, multi-object tracking, facial recognition, ReID, image processing, model deployment and inference, network data analysis and backend development. He has published multiple first-author papers at leading international AI conferences rated CCF-A and is the first-listed inventor on several granted Chinese invention patents. His industry experience includes computer vision for unmanned retail, financial anti-fraud systems, forgery detection, network traffic alerts and backend infrastructure.
Text: SINGAPORE-LAW-PRESS