Promising papers related to DSL in ICML 2026

Spectral Analysis of Model

• Spectral-Informed Neural Networks Outperform Spectral Methods in High-dimensional PDEs

• When Shared Knowledge Hurts: Spectral Over-Accumulation in Model Merging

• SD-MoE: Spectral Decomposition for Effective Expert Specialization

• Geometry of Reason: Spectral Signatures of Valid Mathematical Reasoning

Spectral Generation

• Physics-informed Diffusion Models in Spectral Space

• Primal-Spectral Generative Modeling: Fast Analytical Generation via Pseudoinverse Lévy Inversion

• Revisiting Spectral Representations in Generative Diffusion Models

• Designing noise schedules for diffusion models with spectral analysis

Spectrum-based Applications

• FRIGID: Scaling Diffusion-Based Molecular Generation from Mass Spectra at Training and Inference Time

• SymSpectra: Symmetric Information Bottleneck Framework for Molecular Structure Recognition under Imbalanced Settings

Spectral Optimization

• Fast Spectrally Sparse Signal Reconstruction via Jacobi-Preconditioned Gradient Descent

• Spectra: Rethinking Optimizers for LLMs Under Spectral Anisotropy

• FUSE: Full‑spectrum Unlearnable Examples via Spectral Equalization