data synthesis
Data synthesis refers to the process of generating new data samples from existing datasets, often using techniques like generative models to create realistic and diverse examples that can enhance training or testing datasets.
- Data-Juicer 2.0: Cloud-Scale Adaptive Data Processing for and with Foundation Models
- GRIP: A Graph-Based Reasoning Instruction Producer
- Generating Multi-Table Time Series EHR from Latent Space with Minimal Preprocessing
- Quality-Driven Curation of Remote Sensing Vision-Language Data via Learned Scoring Models
- SCAN: Self-Denoising Monte Carlo Annotation for Robust Process Reward Learning
- Sample-Efficient Multi-Round Generative Data Augmentation for Long-Tail Instance Segmentation
- Semantic-guided Diverse Decoding for Large Language Model
- Training-Free Constrained Generation With Stable Diffusion Models
- TreeSynth: Synthesizing Diverse Data from Scratch via Tree-Guided Subspace Partitioning
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability