representation space
The representation space is a conceptual space in which input data is mapped, enabling models to learn meaningful features or embeddings. The structure of this space affects the learning process and the model's ability to generalize.
- Controlling Thinking Speed in Reasoning Models
- Diffusion Generative Modeling on Lie Group Representations
- Fine-grained List-wise Alignment for Generative Medication Recommendation
- From Pretraining to Pathology: How Noise Leads to Catastrophic Inheritance in Medical Models
- The Dual Nature of Plasticity Loss in Deep Continual Learning: Dissection and Mitigation
- VIPAMIN: Visual Prompt Initialization via Embedding Selection and Subspace Expansion
- WeatherPrompt: Multi-modality Representation Learning for All-Weather Drone Visual Geo-Localization