The paper proves that best-of-K explorative training rewards the number of distinct valid outputs a model can emit ('freedom'), and shows freedom-based selection beats child-validation selection under distribution shift.
Sepp Hochreiter and J¨ urgen Schmidhuber
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
Why the Third Axis Is Freedom
The paper proves that best-of-K explorative training rewards the number of distinct valid outputs a model can emit ('freedom'), and shows freedom-based selection beats child-validation selection under distribution shift.