Search papers, labs, and topics across Lattice.
This paper introduces Claim-Level Reliability Assessment (CLR), a training-free framework that enhances test-time reasoning by reallocating compute resources from solution sampling to targeted verification of decision-critical claims. By focusing on semantic falsification, CLR effectively isolates logical anchors in reasoning traces, allowing for a more efficient identification of errors. The results show that CLR significantly improves performance metrics, such as pass@1 and self-consistency, across multiple large language models and reasoning benchmarks while reducing token usage.
Targeted verification can boost reasoning accuracy by over 27 percentage points while using 37% fewer tokens, transforming how we evaluate model outputs.
We propose claim-level falsification as a principle for test-time scaling and instantiate it through Claim-Level Reliability Assessment (CLR), a training-free framework that reallocates test-time compute from additional solution sampling to targeted verification. Since whole-trace evaluation often obscures decisive errors due to signal dilution from routine tokens, CLR condenses each reasoning trace into a compact set of decision-critical claims, thereby isolating its logical anchors. Furthermore, recognizing the inherent difficulty of generating entirely correct solutions under fixed model capabilities, CLR shifts the focus to semantic falsification. This approach exploits a fundamental asymmetry between solution construction and claim refutation. Constructing a valid solution requires a flawless reasoning path, whereas refuting an incorrect claim requires identifying only a single decisive flaw. This targeted search for negative evidence systematically compresses the survival space of high-confidence incorrect traces, effectively suppressing erroneous consensus via nonlinear reliability scoring. Across four LLMs and four reasoning benchmarks under matched budgets, CLR generally improves upon pass@1 and self-consistency. On GPT-OSS-20B/CMIMC25, for instance, CLR exceeds pass@1 by 27.15 percentage-points and raises self-consistency accuracy from 77.50\% to 82.19\% with 37.0\% fewer tokens.
