claim-assessment

Claim-Level Reliability Assessment (CLR) is a training-free framework that improves reasoning accuracy in LLMs by verifying critical claims instead of sampling more solutions, reducing token use while boosting performance.