Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning

TL;DR

ReASC is a novel method that enhances reasoning reliability in large language models by adaptively sampling responses based on evidence sufficiency, significantly reducing inference costs while maintaining accuracy.

Contribution

It introduces a reliability-aware adaptive sampling framework that improves efficiency by using response confidence and evidence sufficiency instead of count-based rules.

Findings

ReASC achieves the best accuracy-cost trade-off across five models and four datasets.

It reduces inference cost by up to 70% on GSM8K while preserving accuracy.

ReASC improves inference efficiency across model scales from 3B to 27B parameters.

Abstract

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.