LLMs that pass neurosurgery board questions are easily distracted by one irrelevant sentence, with open-source models losing up to 20.4 percent accuracy.
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
CONDITIONAL 1representative citing papers
citing papers explorer
-
Evaluating the performance and fragility of large language models on the self-assessment for neurological surgeons
LLMs that pass neurosurgery board questions are easily distracted by one irrelevant sentence, with open-source models losing up to 20.4 percent accuracy.