Data associated with the paper 'Scaling Text-to-Image Diffusion Transformers with Representation Autoencoders'. The dataset is used for training and evaluating Scale-RAE, a framework investigating scaling Representation Autoencoders for large-scale, freeform text-to-image generation. It includes web, synthetic, and text-rendering data for scaling RAE decoders beyond ImageNet.
Use Cases
- Training text-to-image diffusion transformers based on the described web, synthetic, and text-rendering data.
- Evaluating the scaling performance of Representation Autoencoder (RAE) decoders beyond ImageNet.
- Benchmarking freeform text-to-image generation frameworks based on the multi-source data composition.
Strengths
- Data is associated with a published research paper, suggesting a defined research purpose.
- Includes multiple data sources (web, synthetic, text-rendering) for scaling experiments.
Limitations
- Description metadata is limited; actual data quality requires manual inspection after download.
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
Provenance
- Source
- nyu-visionx
- Collection Method
- Likely collected from web sources, synthetic generation, and text-rendering processes.
- Freshness
- Last updated 2026-01-24 23:14:16; freshness should be verified.