NeurIPS 2021 Reveals Outstanding Paper Awards Amid Record Submissions
NeurIPS 2021 honors six Outstanding Papers alongside new benchmark and Test of Time awards. The prestigious AI conference accepts 2,344 papers from a record-breaking pool of 9,122 submissions.
NeurIPS 2021 announces its top paper awards as the prestigious AI conference prepares to kick off, honoring six submissions with Outstanding Paper Awards alongside two Datasets and Benchmarks Track awards and one Test of Time Award. A record-breaking 9,122 papers face review this year, resulting in 2,344 accepted papers and an acceptance rate of 26 percent, which marks the highest rate since 2013.
Corporate and academic leaders dominate the accepted paper rankings, with Google leading the industry contributors at 177 papers, followed by Microsoft at 116 and DeepMind at 81. On the academic side, MIT leads with 142 accepted papers, closely followed by Stanford University with 139 and Carnegie Mellon University with 117, while international contributions show the United States, China, and the United Kingdom as the top three contributing regions.
The awarded research tackles fundamental AI challenges, such as a Microsoft and Stanford collaboration that proposes a theoretical model explaining why deep networks require excess parameters to achieve robustness. Another recognized DeepMind paper provides a clear exposition on the expressivity of Markov rewards in reinforcement learning, highlighting the rigorous theoretical focus of this year's top honors.