Cross-family verification improves LLM solution quality more than self-verification on math and logic tasks, with a new verifier-gain metric predicting gains and post-training weakening self-improvement while strengthening cross-family gains.
verifier gain
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
-
When Does Verification Pay Off? A Closer Look at LLMs as Solution Verifiers
Cross-family verification improves LLM solution quality more than self-verification on math and logic tasks, with a new verifier-gain metric predicting gains and post-training weakening self-improvement while strengthening cross-family gains.