NeurIPS 2024 Honors Generative Adversarial Networks and Seq2Seq with Test of Time Awards
NeurIPS 2024 recognizes two groundbreaking papers from 2014 that have fundamentally shaped modern artificial intelligence. The awarded works introduce generative adversarial networks and the encoder-decoder architecture that powers today's large language models.
NeurIPS 2024 officially recognizes two groundbreaking papers from the 2014 conference with the prestigious Test of Time Award. Organizers make a rare exception this year to honor two separate works because of their undeniable and massive influence on the entire artificial intelligence landscape. The celebrated papers are "Generative Adversarial Nets" by Ian Goodfellow and colleagues, and "Sequence to Sequence Learning with Neural Networks" by Ilya Sutskever and colleagues.
The Generative Adversarial Nets paper currently boasts over 85,000 citations and serves as a foundational pillar for modern generative modeling. This work inspires countless research advances over the past decade and enables generative AI to impact a diverse range of practical applications across vision data and other domains. Meanwhile, the Sequence to Sequence Learning paper accumulates more than 27,000 citations by establishing the essential encoder-decoder architecture.
This encoder-decoder framework directly paves the way for later attention-based improvements that drive today's rapid advances in large language models and foundation models. Attendees have the opportunity to see both authors present their historically significant works in person at the conference. These presentations take place on Friday, December 13th, 2024, and feature a short question and answer session immediately following the talks.