Theoretical Limits of Language Model Alignment
Imported from official source
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 the highest-reward output among NNN independent samples. …
This version
- Version
- 1 of 1
- Recorded
- November 12, 2026 18:02
- Change
- Initial
- Content hash
7ed4b043f3e6d9e8e5b460b7c060b9bc- All versions
- Revision history