Self-Consistency from Only Two Samples: CoT-PoT Ensembling for Efficient LLM Reasoning

• URL: http://arxiv.org/abs/2604.17433v1
• Date: Sun, 19 Apr 2026 13:26:04 GMT
• Title: Self-Consistency from Only Two Samples: CoT-PoT Ensembling for Efficient LLM Reasoning
• Abstract summary: Self-consistency (SC) is a popular technique for improving the reasoning accuracy of large language models by aggregating multiple sampled outputs.<n>We introduce a hybrid ensembling approach that leverages the complementary strengths of two distinct modes of reasoning: Chain-of-Thought (CoT) and Program-of-Thought (PoT)<n>We show that CoT-PoT ensembling not only improves overall accuracy, but also drastically reduces the number of samples required for SC by a factor of 9.3x.
• Score: 8.541220267844341
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Self-consistency (SC) is a popular technique for improving the reasoning accuracy of large language models by aggregating multiple sampled outputs, but it comes at a high computational cost due to extensive sampling. We introduce a hybrid ensembling approach that leverages the complementary strengths of two distinct modes of reasoning: Chain-of-Thought (CoT) and Program-of-Thought (PoT). We describe a general framework for combining these two forms of reasoning in self-consistency, as well as particular strategies for both full sampling and early-stopping. We show that CoT-PoT ensembling not only improves overall accuracy, but also drastically reduces the number of samples required for SC by a factor of 9.3x. In particular, the majority of tasks (78.6%) can be addressed with only two samples, which has not been possible with any prior SC methods.

Related papers

• QuasiMoTTo: Quasi-Monte Carlo Test-Time Scaling [35.00824891749274]
  Scaling inference compute, by generating many parallel attempts per problem, is a costly but reliable lever for improving language model capabilities.<n>We introduce QuasiMoTTo, which uses correlated samples as a drop-in replacement for i.i.d. samples.<n>Our empirical analysis focuses on understanding how efficiently QuasiMoTTo can turn compute into performance.
  arXiv  Detail & Related papers  (2026-07-01T17:10:08Z)
• Truncated Step-Level Sampling with Process Rewards for Retrieval-Augmented Reasoning [32.295907409325615]
  Training large language models to reason with search engines via reinforcement learning is hindered by a credit assignment problem.<n>We propose SLATE, a framework built on two complementary ideas.<n> Experiments on seven QA benchmarks confirm that SLATE consistently outperforms both sparse-reward and process-reward baselines.
  arXiv  Detail & Related papers  (2026-02-26T19:05:40Z)
• Breaking the Overscaling Curse: Thinking Parallelism Before Parallel Thinking [14.561556728044918]
  In system-level evaluations, a global parallelism level N is allocated to all samples, typically set large to maximize overall dataset accuracy.<n>Some samples can achieve comparable performance with a smaller N' N, causing budget redundancy.<n>We formalize and quantify the overscaling curse, showing its universality and severity in practice, and analyze its trigger mechanism.
  arXiv  Detail & Related papers  (2026-01-29T12:22:45Z)
• Enhancing the Outcome Reward-based RL Training of MLLMs with Self-Consistency Sampling [90.87033586963828]
  Outcome-reward reinforcement learning (RL) is a common and increasingly significant way to refine the step-by-step reasoning of multimodal large language models (MLLMs)<n>We propose Self-Consistency Sampling (SCS) to correct this issue.<n>Based on Qwen2.5-VL-7B-Instruct, SCS improves accuracy by up to 7.7 percentage points on six multimodal benchmarks with negligible extra computation.
  arXiv  Detail & Related papers  (2025-11-13T18:59:57Z)
• CoT-Saliency: Unified Chain-of-Thought Reasoning for Heterogeneous Saliency Tasks [96.64597365827046]
  We present the first unified framework that jointly handles three operationally heterogeneous saliency tasks.<n>We introduce a Chain-of-Thought (CoT) reasoning process in a Vision-Language Model (VLM) to bridge task heterogeneity.<n>We show our model matches or outperforms specialized SOTA methods and strong closed-source VLMs across all tasks.
  arXiv  Detail & Related papers  (2025-11-01T04:37:01Z)
• Reasoning with Sampling: Your Base Model is Smarter Than You Think [52.639108524651846]
  We propose a simple iterative sampling algorithm leveraging the base models' own likelihoods.<n>We show that our algorithm offers substantial boosts in reasoning that nearly match and even outperform those from RL.<n>Our method does not require training, curated datasets, or a verifier.
  arXiv  Detail & Related papers  (2025-10-16T17:18:11Z)
• Fractured Chain-of-Thought Reasoning [61.647243580650446]
  We introduce Fractured Sampling, a unified inference-time strategy that interpolates between full CoT and solution-only sampling.<n>We show that Fractured Sampling consistently achieves superior accuracy-cost trade-offs, yielding steep log-linear scaling gains in Pass@k versus token budget.
  arXiv  Detail & Related papers  (2025-05-19T11:30:41Z)
• Reasoning Aware Self-Consistency: Leveraging Reasoning Paths for Efficient LLM Sampling [9.44858963874474]
  Self-Consistency mitigates hallucinations in Large Language Models (LLMs) by sampling multiple reasoning paths.<n>We introduce Reasoning-Aware Self-Consistency (RASC), a novel framework that enhances sampling efficiency and reasoning faithfulness.
  arXiv  Detail & Related papers  (2024-08-30T05:14:59Z)
• Single-Stage Visual Relationship Learning using Conditional Queries [60.90880759475021]
  TraCQ is a new formulation for scene graph generation that avoids the multi-task learning problem and the entity pair distribution. We employ a DETR-based encoder-decoder conditional queries to significantly reduce the entity label space as well. Experimental results show that TraCQ not only outperforms existing single-stage scene graph generation methods, it also beats many state-of-the-art two-stage methods on the Visual Genome dataset.
  arXiv  Detail & Related papers  (2023-06-09T06:02:01Z)
• Single-Trajectory Distributionally Robust Reinforcement Learning [21.955807398493334]
  We propose Distributionally Robust RL (DRRL) to enhance performance across a range of environments. Existing DRRL algorithms are either model-based or fail to learn from a single sample trajectory. We design a first fully model-free DRRL algorithm, called distributionally robust Q-learning with single trajectory (DRQ)
  arXiv  Detail & Related papers  (2023-01-27T14:08:09Z)
• On-Demand Sampling: Learning Optimally from Multiple Distributions [63.20009081099896]
  Social and real-world considerations have given rise to multi-distribution learning paradigms. We establish the optimal sample complexity of these learning paradigms and give algorithms that meet this sample complexity. Our algorithm design and analysis are enabled by our extensions of online learning techniques for solving zero-sum games.
  arXiv  Detail & Related papers  (2022-10-22T19:07:26Z)
• The Simulator: Understanding Adaptive Sampling in the Moderate-Confidence Regime [52.38455827779212]
  We propose a novel technique for analyzing adaptive sampling called the em Simulator. We prove the first instance-based lower bounds the top-k problem which incorporate the appropriate log-factors. Our new analysis inspires a simple and near-optimal for the best-arm and top-k identification, the first em practical of its kind for the latter problem.
  arXiv  Detail & Related papers  (2017-02-16T23:42:02Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.