NeurIPS 2022 outstanding papers are announced! Stanford University successfully 'defended' the title, and Li Feifei was named on the list of top apprentices

The annual international artificial intelligence summitNeurIPS, the full name of Neural Information Processing Systems (Neural Information Processing Systems), is usually held in December every year.

This year is the 36th NeurIPS, which will last for two weeks starting from November 28: The first week will be held in New Orleans, USA. , and the second week was converted to an online meeting.

As the "appetizer" for the official meeting, as usual, the NeurIPS organizing committee will officially announce the list of winning papers. The three awards are Outstanding Papers Award(Outstanding Papers ), Outstanding Datasets and Benchmarks Papers Award(Outstanding Datasets and Benchmarks Papers) and Time Test Award(Test of Time Award).

As one of the most prestigious artificial intelligence events in the world, NeurIPS received a total of 10,411 papers this year, of which 2,672 were accepted after review, with an acceptance rate of only 25.6%.

And the paper that finally won the award is one of the best, and can fully represent the highest level of current neuroscience and artificial intelligence research.

On the award list, a total of 13 papers won the Outstanding Paper Award this year, the number was last year (6 papers) Twice; Outstanding Dataset and Benchmark Paper Award and Time Test Award were awarded to 2 and 1 papers respectively, and the number was the same as last year.

According to the NeurIPS review, the committee selected these papers because they"have outstanding creativity, insight, clarity, and potential to change the world".

Among the 13 papers that won the Outstanding Paper Award, 3 papers were provided by Chinese teams, and 2 results were completed by the "All-Chinese Class" .

It is worth mentioning that "MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge" won the Outstanding Dataset and Benchmark Paper Award, by It was completed by Linxi Fan (first author) and Yuke Zhu (co-advisor), two disciples of Li Feifei, a Chinese-American academician and professor at Stanford University.

This article proposes a novel agent learning algorithm that can solve problems specified in a free-form language by introducing a new framework MineDojo built on the game "Minecraft". various open-ended tasks.

Among the 16 award-winning papers this year, 4 have the participation of researchers from Stanford University, and in the 2021 selection, they also have 3 selected. In the field of artificial intelligence research, the leading advantage of this top American school is evident.

Finally, the most interesting award selected every year is the Time Test Award, which specializes in selecting papers from ancient times.

Last year this award was won by researchers from Princeton University, this year it was awarded to Alex Krizhevsky, Ilya Sutskever and Geoffrey E. Hinton of the University of Toronto for their work published in 2012 "ImageNet Classification with Deep Convolutional Neural Networks".

Among the reasons for the award, the NeurIPS judges wrote, “As the first CNN to be trained on the ImageNet challenge, this 2012 study Far exceeding the state-of-the-art technology at the time, it opened a new wave of deep learning and had a profound impact on the machine learning community."

Outstanding Paper Award

1. Is Out-of-Distribution Detection Learnable?

##（Zhen Fang, Yixuan Li, Jie Lu, Jiahua Dong, Bo Han, Feng Liu）

2、Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding

（Chitwan Saharia、William Chan、Saurabh Saxena等）

3、Elucidating the Design Space of Diffusion-Based Generative Models

（Tero Karras、Miika Aittala、Timo Aila、Samuli Laine）

4、ProcTHOR: Large-Scale Embodied AI Using Procedural Generation

（Matt Deitke、Eli VanderBilt、Alvaro Herrasti等）

5、Using natural language and program abstractions to instill human inductive biases in machines

（Sreejan Kumar、Carlos G. Correa、Ishita Dasgupta等）

6、A Neural Corpus Indexer for Document Retrieval

（Yujing Wang、Yingyan Hou、Haonan Wang等）

7、High-dimensional limit theorems for SGD: Effective dynamics and critical scaling

（Gerard Ben Arous、Reza Gheissari、Aukosh Jagannath）

8、Gradient Descent: The Ultimate Optimizer

（Kartik Chandra、Audrey Xie、Jonathan Ragan-Kelley等）

9、Riemannian Score-Based Generative Modelling

（Valentin De Bortoli、 Emile Mathieu、Michael John Hutchinson等）

10、Gradient Estimation with Discrete Stein Operators

（Jiaxin Shi、Yuhao Zhou、Jessica Hwang等）

11、An empirical analysis of compute-optimal large language model training

（Jordan Hoffmann、Sebastian Borgeaud、Arthur Mensch等）

12、Beyond neural scaling laws: beating power law scaling via data pruning

（Ben Sorscher、Robert Geirhos、Shashank Shekhar等）

13、On-Demand Sampling: Learning Optimally from Multiple Distributions

（Nika Haghtalab、Michael Jordan、Eric Zhao）

杰出数据集和基准论文奖

1、LAION-5B: An open large-scale dataset for training next generation image-text models

（Christoph Schuhmann 、 Romain Beaumont 、 Richard Vencu等）

2、MineDojo: Building Open-Ended Embodied Agents with Internet-Scale Knowledge

（Linxi Fan、Guanzhi Wang、Yunfan Jiang等）

时间检验奖

1、ImageNet Classification with Deep Convolutional Neural Networks

（Alex Krizhevsky 、 Ilya Sutskever、Geoffrey E. Hinton）

详细获奖名单可见：

https://blog.neurips.cc/2022/11/21/announcing-the-neurips-2022-awards/

The above is the detailed content of NeurIPS 2022 outstanding papers are announced! Stanford University successfully 'defended' the title, and Li Feifei was named on the list of top apprentices. For more information, please follow other related articles on the PHP Chinese website!