Apple AI Updates: April 28, 2026

1. Apple Showcases Dozens of Papers at ICLR 2026, Highlighting Parallel-Trainable Nonlinear RNNs and Smarter Data Pruning

Apple. ICLR 2026 wrapped in Rio de Janeiro on April 27 with Apple presenting dozens of papers and workshops, headlined by “ParaRNN: Unlocking Parallel Training of Nonlinear RNNs for Large Language Models” and “Cram Less to Fit More: Training Data Pruning Improves Memorization of Facts.” The first targets the long-running gap between nonlinear RNN architectures and the GPU efficiency of Transformers; the second pushes back on bigger-corpus-is-always-better, arguing that aggressive pruning can improve factual recall in trained LLMs. The cluster of submissions reinforces Apple’s bias toward on-device-friendly architectures and data efficiency rather than scaling raw parameter counts. Source