A router trained on a self-generated dataset can cut token usage in zero-shot QA by switching among three prompting styles, but the claimed accuracy superiority is not supported by the paper's own numbers.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
DynamicMind: A Tri-Mode Thinking System for Large Language Models
A router trained on a self-generated dataset can cut token usage in zero-shot QA by switching among three prompting styles, but the claimed accuracy superiority is not supported by the paper's own numbers.