1,200 ICLR 2026 Papers with Public Code or Data [R]

PaperDigest has released a significant resource for the AI research community: a curated list of roughly 1,200 papers accepted to the International Conference on Learning Representations (ICLR) 2026 that come with public code, data, or demo links. This list, extracted directly from the paper submissions, accounts for approximately 22% of the over 5,300 total accepted papers. The resource includes direct links to the assets, with a dedicated 'code' column that takes users straight to the GitHub repository or official project site. This provides an unprecedented early look at the practical, reproducible research that will define the next wave of AI advancements, all available months before the conference officially begins in Rio de Janeiro on April 22, 2026.

This early access to code and data is a major accelerant for the field. It allows other researchers, engineers, and developers to immediately validate results, replicate experiments, and integrate new techniques into their own projects without waiting for the formal conference proceedings. The list highlights a growing trend toward reproducibility and open science in AI, though it's noted that some repositories may not be fully public until the conference start. For professionals, this means the 2026 research cycle effectively begins now, with the most implementable work from a premier venue readily available for exploration and application.