Representation Learning | Nandan Kumar Jhahttps://www.nankj.com/tag/representation-learning/Representation LearningWowchemy (https://wowchemy.com)en-us&copy Nandan Kumar Jha 2026Wed, 06 May 2026 00:00:00 +0000https://www.nankj.com/media/icon_hu34d6c9dd6b5727d133d5adaccb31f869_37839_512x512_fill_lanczos_center_2.pngRepresentation Learninghttps://www.nankj.com/tag/representation-learning/Same Architecture, Different Capacity: Optimizer-Induced Spectral Scaling Lawshttps://www.nankj.com/publication/optimizer-induced-spectral-scaling-laws/Wed, 06 May 2026 00:00:00 +0000https://www.nankj.com/publication/optimizer-induced-spectral-scaling-laws/NerVE: Nonlinear Eigenspectrum Dynamics in LLM Feed-Forward Networkshttps://www.nankj.com/publication/nerve/Thu, 23 Apr 2026 00:00:00 +0000https://www.nankj.com/publication/nerve/Spectral Scaling Laws in Language Models: How Effectively Do Feed-Forward Networks Use Their Latent Space?https://www.nankj.com/publication/spectral-scaling-laws/Wed, 01 Oct 2025 00:00:00 +0000https://www.nankj.com/publication/spectral-scaling-laws/<p>Earlier version presented at the <a href="https://icml.cc/virtual/2025/49636" target="_blank" rel="noopener">ICML 2025 Actionable Interpretability Workshop (AIW)</a>.</p>Regularizing the Entropy Landscape of Self-Attention: Towards a Soft Inductive Bias in LLMshttps://www.nankj.com/publication/opt25-entropy-landscape/Mon, 01 Dec 2025 00:00:00 +0000https://www.nankj.com/publication/opt25-entropy-landscape/