Top AI Research Papers of 2020 Show Transformer Expansion and Record Submissions
AI research experiences massive growth in 2020 as top conferences receive record-breaking paper submissions despite virtual formats. Researchers make significant strides by expanding transformer architectures beyond natural language processing into computer vision and protein modeling.
Artificial intelligence research experiences remarkable growth in 2020, with top conferences receiving record-breaking numbers of paper submissions despite shifting to virtual formats. CVPR 2020 receives 6,656 submissions, ICML 2020 sees a 45.7 percent increase to 4,990 papers, and NeurIPS 2020 processes 9,467 submissions, reflecting a 40 percent jump from the previous year. This surge highlights a thriving global research community that remains highly active even during a global pandemic.
A major trend this year involves the successful application of transformer architectures beyond their traditional natural language processing tasks. Researchers bring these powerful networks into computer vision for object detection and panoptic segmentation, as well as into the sciences for protein sequence modeling. Additionally, unsupervised and self-supervised learning methods evolve significantly, establishing themselves as viable alternatives to traditional supervised learning approaches.
Among the standout papers gaining extraordinary attention is "WinoGrande: An Adversarial Winograd Schema Challenge at Scale," which earns an Outstanding Paper Award at AAAI 2020 for advancing common-sense reasoning. Another highly acclaimed study, "Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild," takes home the Best Paper Award at CVPR 2020. These highlighted works represent the cutting-edge innovations that define the AI landscape this year.