A new 3,008-question assertion-reason benchmark shows language models often use semantic similarity as a substitute for causal explanation, with best models reaching only 0.55 MCC.
Malalgoqa: Pedagogical evaluation of counterfactual reasoning in large language models and implications for ai in education
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
dataset 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
dataset 1polarities
unclear 1representative citing papers
citing papers explorer
-
CLEAR-3K: Assessing Causal Explanatory Capabilities in Language Models
A new 3,008-question assertion-reason benchmark shows language models often use semantic similarity as a substitute for causal explanation, with best models reaching only 0.55 MCC.