Safety4h ago
Safety
Consistency Tested as Measure of Philosophical Competence
Researcher Caspar Oesterheld is testing whether consistency across different prompts can quantify a model's conceptual abilities. Preliminary experiments suggest this metric could serve as a future training target. The work aims to improve how language models handle complex philosophical reasoning. A full paper detailing these results is currently in development.