The fourteenth International Conference on Learning Representation (ICLR) concludes today in Rio de Janeiro, capping off nearly a week of presentations, discussions, and research demonstrations by leading AI scientists from academia and the tech industry, including Apple.
See also: When is the next Apple Event?
Although ICLR may not be widely known to the general public, it has been considered for more than a decade as one of the most recognized and respected conferences in machine learning.
This year, the conference took place from April 23 to 27, occupying four pavilions and the conference center of the Riocentro Convention Center in Rio de Janeiro. ICLR 2026 attracted experts and researchers in machine learning and artificial intelligence from around the world, including Yann LeCun from AMI Labs.
It also brought together major technology companies as sponsors and exhibitors, such as Amazon, Tencent, Google, Microsoft, Ant Group, ByteDance, Huawei, Meta, Salesforce, Shopify and Apple, as well as companies from Wall Street and the broader financial industry, such as Capital One, Jane Street, Citadel and others.
See also: Apple's WWDC 2026 starts on June 8
As Apple announced a few days ago, the company had a booth at the conference, where it showcased an impressive open-source model that converts 2D images into 3D spaces in a matter of seconds, as well as running LLM on MLX, Apple's open-source framework for machine learning tasks running on Apple Silicon.
Apple’s booth also served as a recruiting center, with iPads set up so attendees could scan QR codes and apply for machine learning jobs on the spot. Most companies on the show floor were also using the conference as a recruiting pipeline for AI talent.
ICLR featured huge poster areas where researchers presented their work and answered questions about their studies. Apple presented dozens of papers during the conference.
See also: NVIDIA GTC 2026 opens today
Apple also delivered presentations and workshops on some of the papers accepted into the conference, including “ParaRNN: Unleashing Parallel Training of Nonlinear RNNs for Large Language Models,” presented by Federico Danieli, and “Cram Less to Fit More: Pruning Training Data Improves Fact Recall,” presented by Kunal Talwar.
