self-supervised framework
A learning paradigm where models are trained on tasks that don't require labeled data, using inherent structure in the data itself. This approach often leverages pretext tasks to learn meaningful representations.
- $\textit{HiMaCon:}$ Discovering Hierarchical Manipulation Concepts from Unlabeled Multi-Modal Data
- A compressive-expressive communication framework for compositional representations
- C3PO: Optimized Large Language Model Cascades with Probabilistic Cost Constraints for Reasoning
- Jasmine: Harnessing Diffusion Prior for Self-supervised Depth Estimation
- Know Thyself by Knowing Others: Learning Neuron Identity from Population Context
- MoE-Gyro: Self-Supervised Over-Range Reconstruction and Denoising for MEMS Gyroscopes
- MoPFormer: Motion-Primitive Transformer for Wearable-Sensor Activity Recognition
- Spatially-aware Weights Tokenization for NeRF-Language Models
- TRACE: Contrastive learning for multi-trial time series data in neuroscience
- VCM: Vision Concept Modeling with Adaptive Vision Token Compression via Instruction Fine-Tuning