2 papers
cs.LG2024
Aligning language models with human preferences
Tomasz Korbak
Language models (LMs) trained on vast quantities of text data can acquire sophisticated skills such as generating summaries, answering questions or generating code. However, they a…
cs.CL2023
Compositional preference models for aligning LMs
Dongyoung Go, Tomasz Korbak, Germán Kruszewski +2
As language models (LMs) become more capable, it is increasingly important to align them with human preferences. However, the dominant paradigm for training Preference Models (PMs)…