scaling behavior
The relationship between the performance of an AI model and the size of the dataset or computational resources used, crucial for understanding how well models generalize with increased scale.
- Chain-of-Retrieval Augmented Generation
- Grids Often Outperform Implicit Neural Representation at Compressing Dense Signals
- Horizon Reduction Makes RL Scalable
- On the Entropy Calibration of Language Models
- Scaling Off-Policy Reinforcement Learning with Batch and Weight Normalization
- Scaling Offline RL via Efficient and Expressive Shortcut Models