Self-Consistency improves reasoning reliability through multi-sample aggregation, but incurs substantial inference cost. Adaptive self-consistency methods mitigate this issue by adjusting the sampling budget; however, they rely on count-based stopping rules that treat all responses equally, often leading to unnecessary sampling. We propose Reliability-Aware Adaptive Self-Consistency (ReASC), which addresses this limitation by reframing adaptive sampling from response counting to evidence sufficiency, leveraging response-level confidence for principled information aggregation. ReASC operates in two stages: a single-sample decision stage that resolves instances confidently answerable from a single response, and a reliability-aware accumulation stage that aggregates responses by jointly leveraging their frequency and confidence. Across five models and four datasets, ReASC consistently achieves the best accuracy-cost trade-off compared to existing baselines, yielding improved inference efficiency across model scales from 3B to 27B parameters. As a concrete example, ReASC reduces inference cost by up to 70\% relative to self-consistency while preserving accuracy on GSM8K using Gemma-3-4B-it.

翻译：自洽性通过多样本聚合提升推理可靠性，但会产生显著的推理开销。自适应自洽性方法通过调整采样预算缓解此问题；然而，这些方法依赖基于计数的停止规则，平等对待所有响应，常导致不必要的采样。本文提出可靠性感知自适应自洽性方法，通过将自适应采样框架从响应计数重构为证据充分性判定，并利用响应级置信度进行原则性信息聚合，以解决上述局限。ReASC 分两阶段运行：单样本决策阶段解析可仅凭单次响应确信回答的实例，以及可靠性感知累积阶段通过联合利用响应频次与置信度进行响应聚合。在五个模型和四个数据集上的实验表明，与现有基线相比，ReASC 始终实现最佳精度-成本权衡，在参数量从 3B 到 27B 的模型规模上均提升了推理效率。具体而言，在使用 Gemma-3-4B-it 模型处理 GSM8K 数据集时，ReASC 在保持精度的同时，相较标准自洽性方法最高可降低 70% 的推理成本。