A model's accuracy on questions it generates about its own explanations correlates with MMLU-Pro only at r = 0.361, and the paper claims this can serve as a test-set-free ranking proxy.
Introducing claude: Anthropic's ai assistant
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Explain-Query-Test: Self-Evaluating LLMs Via Explanation and Comprehension Discrepancy
A model's accuracy on questions it generates about its own explanations correlates with MMLU-Pro only at r = 0.361, and the paper claims this can serve as a test-set-free ranking proxy.