NCV: A Node-Wise Consistency Verification Approach for Low-Cost Structured Error Localization in LLM Reasoning
This addresses the challenge of imprecise error localization and high token costs in LLM reasoning verification, offering a scalable solution for researchers and practitioners, though it is incremental as it builds on existing verification approaches.
The paper tackled the problem of verifying multi-step reasoning in large language models by introducing Node-wise Consistency Verification (NCV), a training-free framework that localizes errors through lightweight binary checks at the node level, achieving a 10% to 25% improvement in F1 scores and using 6× to 58× fewer tokens than traditional methods.
Verifying multi-step reasoning in large language models is difficult due to imprecise error localization and high token costs. Existing methods either assess entire reasoning chains, suffering attention dilution, or rely on expensive multi-sampling. We introduce Node-wise Consistency Verification (NCV), a training-free framework that recasts verification as lightweight binary consistency checks at the node level. By decomposing the chain of thought into interconnected verification nodes, NCV precisely localizes errors and avoids unnecessary long-form generation. Experiments demonstrate that our approach enhances interpretability and efficiency, presenting a scalable solution for reliable LLM reasoning verification. On public datasets, NCV achieves a 10\% to 25\% improvement in F1 scores over baselines while utilizing $6\times$~$58\times$ fewer tokens than traditional methods like CoT-based verifiers.