A new four-variant benchmark of simple geometric logic questions shows most LLMs score near chance, with performance largely stable across small language variations.
Attention is all you need,
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
-
Reasoning Capabilities and Invariability of Large Language Models
A new four-variant benchmark of simple geometric logic questions shows most LLMs score near chance, with performance largely stable across small language variations.