A new evaluation framework shows that even the best tested LLM only reliably adjusts response complexity in the intended direction 46% of the time across 98 scientific queries.
Journal of Clinical Nursing , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Explain Like I'm 5 or Whatever I Choose: Evaluating the Interactive Potential of Language Model Responses
A new evaluation framework shows that even the best tested LLM only reliably adjusts response complexity in the intended direction 46% of the time across 98 scientific queries.