The NeurIPS Test of Time Award for Tomáš Mikolov and his team for the revolutionary research on language models


A major Test of Time award was presented to researcher Tomas Mikolov and his team at the NeurIPS conference for their revolutionary work that has significantly influenced the shape of today’s large-scale language models. The paper, titled „Distributed Representations of Words and Phrases and their Compositionality,“ which originated back in 2013, has garnered more than 40,000 citations over the years, and this work laid the groundwork for the vector representation of words known as word2vec.

The award was based on a „test of time“, evaluating the success of the work ten years after its publication. The introduction of word2vec in this work demonstrated the ability to learn from large amounts of unstructured text and ushered in a new era in natural language processing. Tomas Mikolov, team leader of the Basic AI Research team at CIIRC, received the prestigious Test of Time Award for his significant contribution to the field of artificial intelligence. His contribution not only influenced the development of language models, but also laid the foundations for the emergence of models including ChatGPT.

Mikolov himself said of the award: „Yesterday at NeurIPS we received the Test of Time award for the word2vec entry from ten years ago. I am really happy about it! I think it’s the first „the best paper“ award I’ve ever received. In fact, the original word2vec paper was rejected at the first ICLR conference in 2013, which got me thinking about how difficult it is for reviewers to predict the future impact of research papers.“

„Without this key work, the development of large-scale neural language models would certainly be slower. Ultimately, word2vec was groundbreaking in its time for several reasons – not just that it was an algorithm that could quickly process large amounts of text data. The reproducibility of the results was also very important, thanks to the release of an open-source implementation. Last but not least, word2vec demonstrated the surprising properties of learned representations that could be used to solve equations such as „KING – MAN + WOMAN (the result was a QUEEN),“ says Tomas Mikolov.

We heartily congratulate Tomas Mikolov and his co-authors on this important recognition of their work!

Source: 1) Announcing the NeurIPS 2023 Paper Awards 


Another media outputs:

1) Fokus Václava Moravce: Vzestup a pád inteligence

2) Podcast with Tomas Mikolov: Elon Musk thinks first of all about money

3) – in CZ

Previous articleCACORR 2024 - scientists present the best of their research
Next articleJan Šedivý commented on the report about the "fake" Joe Biden, Newsroom ČT24