AI Radar
Top stories
Importance 2 of 5Covered by 1 source

Theoretical Limits of Language Model Alignment

Language model (LM) alignment improves model outputs to reflect human preferences while preserving the capabilities of the base model. The most common alignment approaches are (i) reinforcement learning, which maximizes the expected reward under a KL-divergence constraint, and (ii) best-of-NNN alignment, which selects…