Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911605251375104 |
|---|---|
| author | Kim, Junseok Yang, Nakyeong Min, Kyungmin Jung, Kyomin |
| author_facet | Kim, Junseok Yang, Nakyeong Min, Kyungmin Jung, Kyomin |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_02970 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning Kim, Junseok Yang, Nakyeong Min, Kyungmin Jung, Kyomin Computation and Language Machine Learning 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. |
| title | Reliability-Aware Adaptive Self-Consistency for Efficient Sampling in LLM Reasoning |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2601.02970 |