Iclr_2026

Our works:

• Scaling Laws and Spectra of Shallow Neural Networks in the Feature Learning Regime (Oral)
• Fast Escape, Slow Convergence: Learning Dynamics of Phase Retrieval under Power-Law Data (Oral)
• Statistical Advantage of Softmax Attention: Insights from Single-Location Regression

were accepted for ICLR 2026, two of them as orals! Come see the talks and posters in Rio de Janeiro!