Lossfunk

Future Is Unevenly Distributed: Forecasting Ability of LLMs Depends on What We're Asking

The Sequential Edge: Inverse-Entropy Voting Beats Parallel Self-Consistency at Matched Compute