LLMs recover intended developmental labels accurately on simulated personas but show only fair agreement with humans on real DSCT responses, while default model outputs exhibit stable stage differences favoring larger models.
Justification: Our human annotation is innocuous and thus does not require IRB approval
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Evaluating Developmental Cognition Capabilities of LLMs
LLMs recover intended developmental labels accurately on simulated personas but show only fair agreement with humans on real DSCT responses, while default model outputs exhibit stable stage differences favoring larger models.