Stanford AI Lab Showcases Broad Research at Virtual NeurIPS 2021
The Stanford AI Lab presents a diverse collection of papers and workshops at the virtual NeurIPS 2021 conference. The research covers topics ranging from emergent communication to data pruning and sequence modeling.
The thirty-fifth Conference on Neural Information Processing Systems (NeurIPS) 2021 takes place virtually from December 6th to 14th, featuring a strong presence from the Stanford AI Lab (SAIL). Researchers from SAIL present their latest work across the main conference, the Datasets and Benchmarks track, and various specialized workshops. The lab provides direct links to papers, videos, and supplementary blogs for attendees to explore these advancements.
The SAIL contributions at the main conference highlight several critical areas of artificial intelligence research. Notable projects include efforts to improve the compositionality of neural networks, reverse engineering recurrent neural networks for better interpretability, and combining recurrent, convolutional, and continuous-time models using linear state space layers. Additional papers explore generative transformers for scene synthesis, emergent communication among AI agents, and training deep learning models on reduced datasets through early data pruning.
Beyond presenting research papers, members of the SAIL community actively participate as co-organizers for multiple exciting workshops scheduled for December 13th and 14th. The lab encourages interested individuals to reach out directly to the contact authors and workshop organizers to dive deeper into the innovative work happening at Stanford.