MILES: Modular Instruction Memory with Learnable Selection for Self-Improving LLM Reasoning Step-wise modular memory with learnable, correctness-oriented composition for frozen LLM reasoning. Little by Little: Continual Learning via Incremental Mixture of Rank-1 Associative Memory Experts MoRAM — rank-1 parametric memory as a dynamic mixture-of-experts for continual learning of LLMs. HyperAlign: Learning Hypernetwork for Efficient Test-Time Alignment of Diffusion Models Armotized efficient test-time alignment; End-to-end post-training hypernetworks; Dynamically generating LoRAs. MotionEcho: Restoring Motion Control in Fast (Distilled) Video Generators Test-time distillation that restores generative trajectories broken by step-distilled video models. DyMoE: Continual Learning Dynamic MoE with Token-Level Guidance Dynamic mixture-of-experts (MoE); Incrementing MoE for model expansion; Token assignment guidance for mitigating routing-drift & forgetting; continual learning of large vision-language models. DyMO: Test-time Diffusion Model Alignment Training-free diffusion model alignment with dynamic multi-objective scheduling FlashMo: Scalable Efficient Motion Generation Geometric interpolants and frequency-aware sparsity for scalable efficient motion generation Let Your Image Move with Your Motion! FlexiMMT - Implicit Multi-Object Multi-Motion Transfer for Video Generation Let Your Video Listen to Your Music! Beat-aligned, content-preserving video editing with arbitrary music (MVAA) Mining Your Own Secrets: Continual Learning for T2I Diffusion Models Diffusion classifier scores for continual personalization of text-to-image diffusion models Continual Learning with Bayesian Sparse Neural Networks Learning Bayesian Sparse Networks with Full Experience Replay for Continual Learning Memorizing Normality to Detect Anomaly: Memory-augmented Deep Autoencoder (MemAE) for Unsupervised Anomaly Detection MemAE - Memory-augmented Deep Autoencoder for Unsupervised Anomaly Detection Memory-augmented Dynamic Neural Relational Inference Memory-augmented neural relational inference for dynamical systems From Motion Blur to Motion Flow: a Deep Learning Solution for Removing Heterogeneous Motion Blur A deep learning solution for removing heterogeneous motion blur via dense motion flow estimation Attention-guided Network for Ghost-free High Dynamic Range (HDR) Imaging AHDR - Attention-guided Network for Ghost-free High Dynamic Range Imaging Multi-frame Depth Estimation in Dynamic Scenes Multi-frame depth estimation with dynamic scene understanding MPGL: An Efficient Matching Pursuit Method for Generalized LASSO An efficient matching pursuit method for generalized LASSO
