Think in Parallel, Answer as One: Logit Averaging for Open-Ended Reasoning

• URL: http://arxiv.org/abs/2512.02874v1
• Date: Tue, 02 Dec 2025 15:35:31 GMT
• Title: Think in Parallel, Answer as One: Logit Averaging for Open-Ended Reasoning
• Authors: Haonan Wang, Chao Du, Kenji Kawaguchi, Tianyu Pang,
• Abstract summary: ThinkMerge is a training-free, plug-and-play decoding strategy.<n>It runs K parallel reasoning traces and averages their next-token logits at synchronization points to produce a single coherent output.
• Score: 102.13989674248116
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Majority voting has proven effective for close-ended question answering by aggregating parallel reasoning traces. However, it is not directly applicable to open-ended reasoning, such as code generation and web-based deep research, where a "majority" over complete solutions is ill-defined. We introduce ThinkMerge, a training-free, plug-and-play decoding strategy that runs K parallel reasoning traces and averages their next-token logits at synchronization points to produce a single coherent output. ThinkMerge integrates seamlessly with vLLM/SGLang and remains compatible with standard decoding techniques such as Top-p/Top-k. Empirically, it matches or surpasses majority voting on AIME and GPQA, while delivering consistent gains on open-ended coding tasks: on LiveCodeBench (hard), pass@1 improves by +8.28% for DeepCoder-14B-Preview and +7.58% for Qwen3-8B. Beyond code, we further show that ThinkMerge improves web-based deep-research agents (e.g., WebSailor-7B/32B) across GAIA, BrowseComp-en/zh, and XbenchDeepSearch. These results demonstrate that parallel test-time scaling can benefit open-ended reasoning without relying on voting over complete outputs.

Related papers

• Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching [66.39914384073145]
  We propose a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates.<n>We find that step-level recombination is most beneficial on harder problems.<n>Our training-free framework improves average accuracy by up to 2 across six math and coding tasks.
  arXiv  Detail & Related papers  (2026-02-26T11:08:39Z)
• LogitsCoder: Towards Efficient Chain-of-Thought Path Search via Logits Preference Decoding for Code Generation [86.08600027874662]
  We propose LogitsCoder, a novel framework that enhances chain-of-thought reasoning through lightweight, logit-level control mechanisms for code generation.<n>We show that LogitsCoder produces more efficient and higher-quality reasoning chains, leading to superior code generation performance compared to baseline methods.
  arXiv  Detail & Related papers  (2026-02-15T08:52:19Z)
• Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probing [76.48164395646019]
  Parallel-Probe is a training-free controller designed to optimize online parallel thinking.<n>It reduces sequential tokens by up to $textbf35.8$% and total token cost by over $textbf25.8$% while maintaining competitive accuracy.
  arXiv  Detail & Related papers  (2026-02-03T18:59:41Z)
• No Compute Left Behind: Rethinking Reasoning and Sampling with Masked Diffusion Models [42.158430350703505]
  Masked diffusion language models are trained to in-fill positions in randomly masked sequences.<n>We propose reasoning-as-infilling and multi-token entropy decoding.<n>Our work demonstrates that the training and compute used by MDLMs unlock many new inference and post-training methods.
  arXiv  Detail & Related papers  (2025-10-22T19:41:27Z)
• DeepPrune: Parallel Scaling without Inter-trace Redundancy [53.62015294143274]
  Over 80% of parallel reasoning traces yield identical final answers, representing substantial wasted computation.<n>We propose DeepPrune, a novel framework that enables efficient parallel scaling through dynamic pruning.<n>Our work establishes a new standard for efficient parallel reasoning, making high-performance reasoning more efficient.
  arXiv  Detail & Related papers  (2025-10-09T17:24:54Z)
• dParallel: Learnable Parallel Decoding for dLLMs [77.24184219948337]
  Diffusion large language models (dLLMs) offer parallel token prediction and lower inference latency.<n>Existing open-source models still require nearly token-length decoding steps to ensure performance.<n>We introduce dParallel, a simple and effective method that unlocks the inherent parallelism of dLLMs for fast sampling.
  arXiv  Detail & Related papers  (2025-09-30T16:32:52Z)
• Eigen-1: Adaptive Multi-Agent Refinement with Monitor-Based RAG for Scientific Reasoning [53.45095336430027]
  We develop a unified framework that combines implicit retrieval and structured collaboration.<n>On Humanity's Last Exam (HLE) Bio/Chem Gold, our framework achieves 48.3% accuracy.<n>Results on SuperGPQA and TRQA confirm robustness across domains.
  arXiv  Detail & Related papers  (2025-09-25T14:05:55Z)
• Reasoning Models Can Be Effective Without Thinking [45.411955744222524]
  We find that bypassing the thinking process via simple prompting, denoted as NoThinking, can be surprisingly effective.<n>Our method outperforms a range of baselines with similar latency using Thinking, and is comparable to Thinking with significantly longer latency (up to 9x)
  arXiv  Detail & Related papers  (2025-04-14T04:08:16Z)
• Z1: Efficient Test-time Scaling with Code [26.374317704720234]
  Large Language Models (LLMs) can achieve enhanced complex problem-solving through test-time computing scaling.<n>We propose an efficient test-time scaling method that trains LLMs on code-related reasoning trajectories.<n>We present a novel Shifted Thinking Window to mitigate overthinking overhead.
  arXiv  Detail & Related papers  (2025-04-01T14:01:50Z)
• Self-Evaluation Guided Beam Search for Reasoning [61.523627290397556]
  We introduce a stepwise self-evaluation mechanism to guide and calibrate the reasoning process of Large Language Model (LLM) We propose a decoding algorithm integrating the self-evaluation guidance via beam search. Our approach surpasses the corresponding Codex-backboned baselines in few-shot accuracy by $6.34%$, $9.56%$, and $5.46%$ on the GSM8K, AQuA, and StrategyQA.
  arXiv  Detail & Related papers  (2023-05-01T02:37:59Z)
• Multi-hop Question Generation with Graph Convolutional Network [58.31752179830959]
  Multi-hop Question Generation (QG) aims to generate answer-related questions by aggregating and reasoning over multiple scattered evidence from different paragraphs. We propose Multi-Hop volution Fusion Network for Question Generation (MulQG), which does context encoding in multiple hops. Our proposed model is able to generate fluent questions with high completeness and outperforms the strongest baseline by 20.8% in the multi-hop evaluation.
  arXiv  Detail & Related papers  (2020-10-19T06:15:36Z)

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.