Mehrtash Harandi

DTO-KD: Dynamic Trade-off Optimization for Effective Knowledge Distillation

Massive Activations are the Key to Local Detail Synthesis in Diffusion Transformers

Modality Alignment across Trees on Heterogeneous Hyperbolic Manifolds

Antibody: Strengthening Defense Against Harmful Fine-Tuning for Large Language Models via Attenuating Harmful Gradient Influence

Align-SAM: Seeking Flatter Minima for Better Cross-Subset Alignment

Sharpness-Aware Minimization in Logit Space Efficiently Enhances Direct Preference Optimization

K-Gen: Unlocking High-Resolution Data-Free Knowledge Distillation via Key Region Generation

Free Lunch at Inference: Test-Time Refinement for Diffusion Models

Unlearn In a Blink

Unleashing Diffusion Transformers for Visual Correspondence by Modulating Massive Activations

Unveiling m-Sharpness Through the Structure of Stochastic Gradient Noise

Geometry-Aware Collaborative Multi-Solutions Optimizer for Model Fine-Tuning with Parameter Efficiency

Fast Feedforward 3D Gaussian Splatting Compression

Token-Level Self-Play with Importance-Aware Guidance for Large Language Models

Optimization on Manifolds with Riemannian Jacobian Regularization

Geometry-aware Distance Measure for Diverse Hierarchical Structures in Hyperbolic Spaces

EraseDiff: Erasing Data Influence in Diffusion Models

SUN: Training-free Machine Unlearning via Subspace

Unified Perspectives on Signal-to-Noise Diffusion Models

Understanding Domain Generalization: A View of Necessity and Sufficiency