mode collapse
Mode collapse is a phenomenon in generative models, where the model tends to produce a limited variety of outputs, ignoring much of the diversity in the training data. This can diminish the overall quality of generated results.
- A Geometry-Aware Metric for Mode Collapse in Time Series Generative Models
- A solvable model of learning generative diffusion: theory and insights
- Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)
- Artificial Hivemind: The Open-Ended Homogeneity of Language Models (and Beyond)
- On scalable and efficient training of diffusion samplers
- PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis