Total 16 · Clear filters
On 21 July 2025, Google DeepMind announced that an advanced Gemini model with Deep Think reached gold-medal standard at the 2025 International Mathematical Olympiad, solving five of six problems for 35 of 42 points under officially certified grading.
On December 20, 2024, the ARC Prize Foundation reported that OpenAI's o3 scored 75.7% on the semi-private ARC-AGI-1 evaluation set, and 87.5% with a high-compute configuration — a jump far beyond earlier models on a benchmark built to resist memorization.
On September 12, 2024, OpenAI introduced o1 and reported that it averaged 83% on AIME 2024 when 64 samples were combined, versus 12% for GPT-4o, plus a Codeforces Elo of 1673.
On 25 July 2024, Google DeepMind announced that AlphaProof and AlphaGeometry 2 solved four of the six 2024 International Mathematical Olympiad problems, scoring 28 of 42 points. The company said this reached silver-medal level, just below the 29-point gold threshold.
On 30 November 2020 DeepMind announced that the organisers of CASP14 had recognised its AlphaFold system as a solution to the 50-year-old protein-folding problem, with accuracy approaching that of experimental methods.
In December 2018 DeepMind's AlphaFold placed first overall at CASP13, a blind protein-structure prediction assessment, giving the best prediction for 25 of the 43 hardest targets.
The 2015 ImageNet classification contest was won by ResNet, a network up to 152 layers deep whose layers learn residual functions relative to their inputs; it reached 3.57% error on the test set and, by one account, exceeded human performance.
The DARPA Robotics Challenge Finals, held on 5 and 6 June 2015 at the Fairplex in Pomona, California, put 23 teams and their robots through eight tasks; Team KAIST's DRC-Hubo took first place and the $2 million prize.
At ILSVRC 2014, GoogLeNet won the classification track with 6.67% top-5 error, while VGG's network won localisation and took second in classification; more than 50 institutions took part.
A two-day DARPA competition in December 2013 in which 16 teams put prototype disaster-response robots through eight simulated tasks; SCHAFT of Tokyo scored highest with 27 points out of a possible 32.
In 2013 the ILSVRC classification task was won by Clarifai with an ensemble of convolutional networks at 11.7% top-5 error, while OverFeat won localisation. Zeiler and Fergus's visualization paper, also reporting ImageNet results, appeared the same month.
In October 2012 AlexNet, a deep neural network built by a University of Toronto team, won the ImageNet image-classification challenge by a wide margin. Its result showed that large neural networks trained on GPUs could outperform hand-designed vision methods.
A Kaggle competition run in August–October 2012 in which competitors predicted molecules' biological activities from numerical descriptors of their chemical structures; a deep-learning team led by Geoffrey Hinton took first place.
On 13 April 2010, Fei-Fei Li announced the first ImageNet Large Scale Visual Recognition Challenge, a 1,000-category image categorization benchmark. It ran for eight editions, from 2010 to 2017.
At ICDAR 2009, Alex Graves won three connected handwriting competitions — French, Farsi and Arabic — with LSTM networks trained using Connectionist Temporal Classification, the first time a recurrent neural network won an international contest.
An open competition run by Netflix from 2006 to 2009 that offered $1,000,000 to any team able to improve its movie recommendation algorithm by 10%, and released 100 million anonymous ratings for contestants to work with.
