Why evaluation metrics reward the wrong behaviour

Popularity Index: 10 Updated at 2026-10-10 09:19 UTC

Sites

Related Topics

Catastrophic forgetting when you fine-tune on new data Data hunger: what happens when the web runs out of usable text The long tail of rare cases that no benchmark captures Sample efficiency: why a model needs a million examples for one skill How much of a generative model's knowledge is memorization rather than generalization? Model collapse is what happens when generative models train on their own output What happens to a generative model when its training data runs out? Synthetic data solved the scarcity problem and created a new one Scaling laws predicted the curve and nobody predicted the data wall Why are so many AI papers impossible to reproduce?

Hot Categories

Featured Categories