Transformers are confined to descriptive uncertainty, with token entropy remaining constant across tasks despite large accuracy differences, implying that regulatory uncertainty requires thermodynamic coupling between error and energy cost.
From heuristics to understanding: how neuromorphic autonomy enables true world models
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Descriptive versus Regulatory Uncertainty in Bounded Predictive Systems
Transformers are confined to descriptive uncertainty, with token entropy remaining constant across tasks despite large accuracy differences, implying that regulatory uncertainty requires thermodynamic coupling between error and energy cost.