Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching

• URL: http://arxiv.org/abs/2602.22871v1
• Date: Thu, 26 Feb 2026 11:08:39 GMT
• Title: Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching
• Authors: Roy Miles, Aysim Toker, Andreea-Maria Oncescu, Songcen Xu, Jiankang Deng, Ismail Elezi,
• Abstract summary: 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.
• Score: 66.39914384073145
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reasoning with large language models often benefits from generating multiple chains-of-thought, but existing aggregation strategies are typically trajectory-level (e.g., selecting the best trace or voting on the final answer), discarding useful intermediate work from partial or "nearly correct" attempts. We propose Stitching Noisy Diffusion Thoughts, a self-consistency framework that turns cheap diffusion-sampled reasoning into a reusable pool of step-level candidates. Given a problem, we (i) sample many diverse, low-cost reasoning trajectories using a masked diffusion language model, (ii) score every intermediate step with an off-the-shelf process reward model (PRM), and (iii) stitch these highest-quality steps across trajectories into a composite rationale. This rationale then conditions an autoregressive (AR) model (solver) to recompute only the final answer. This modular pipeline separates exploration (diffusion) from evaluation and solution synthesis, avoiding monolithic unified hybrids while preserving broad search. Across math reasoning benchmarks, we find that step-level recombination is most beneficial on harder problems, and ablations highlight the importance of the final AR solver in converting stitched but imperfect rationales into accurate answers. Using low-confidence diffusion sampling with parallel, independent rollouts, our training-free framework improves average accuracy by up to 23.8% across six math and coding tasks. At the same time, it achieves up to a 1.8x latency reduction relative to both traditional diffusion models (e.g., Dream, LLaDA) and unified architectures (e.g., TiDAR). Code is available at https://github.com/roymiles/diffusion-stitching.

Related papers

• Free Lunch for Pass@$k$? Low Cost Diverse Sampling for Diffusion Language Models [17.37935640125399]
  We propose a training free, low cost intervention to enhance generative diversity in Diffusion Language Models.<n>Our approach modifies intermediate samples in a batch sequentially, where each sample is repelled from the feature space of previous samples.<n>Unlike prior methods that require retraining or beam search, our strategy incurs negligible computational overhead.
  arXiv  Detail & Related papers  (2026-03-05T07:35:07Z)
• Self-Rewarding Sequential Monte Carlo for Masked Diffusion Language Models [58.946955321428845]
  This work presents self-rewarding sequential Monte Carlo (SMC)<n>Our algorithm stems from the observation that most existing MDLMs rely on a confidence-based sampling strategy.<n>We introduce the trajectory-level confidence as a self-rewarding signal for assigning particle importance weights.
  arXiv  Detail & Related papers  (2026-02-02T09:21:45Z)
• Efficiency vs. Fidelity: A Comparative Analysis of Diffusion Probabilistic Models and Flow Matching on Low-Resource Hardware [0.0]
  Denoising Diffusion Probabilistic Models (DDPMs) have established a new state-of-the-art in generative image synthesis.<n>This study presents a comparative analysis of DDPMs against the emerging Flow Matching paradigm.
  arXiv  Detail & Related papers  (2025-11-24T18:19:42Z)
• RFG: Test-Time Scaling for Diffusion Large Language Model Reasoning with Reward-Free Guidance [101.30279597148973]
  We propose reward-free guidance (RFG) for guiding the reasoning trajectory of dLLMs without explicit process reward.<n>RFG consistently yields significant improvements across all tasks and model types, achieving accuracy gains of up to 9.2%.
  arXiv  Detail & Related papers  (2025-09-29T23:59:16Z)
• Accelerating LLM Reasoning via Early Rejection with Partial Reward Modeling [12.835376812101323]
  We introduce the hypothesis that PRMs are also Partial Reward Models.<n>This allows for principled early rejection based on intermediate token-level signals.<n>On math reasoning benchmarks, our method achieves up to 1.4$times$-9$times$ reduction in inference FLOPs without degrading final performance.
  arXiv  Detail & Related papers  (2025-08-04T00:58:56Z)
• Inference-Time Scaling of Diffusion Language Models with Particle Gibbs Sampling [70.8832906871441]
  We study how to steer generation toward desired rewards without retraining the models.<n>Prior methods typically resample or filter within a single denoising trajectory, optimizing rewards step-by-step without trajectory-level refinement.<n>We introduce particle Gibbs sampling for diffusion language models (PG-DLM), a novel inference-time algorithm enabling trajectory-level refinement while preserving generation perplexity.
  arXiv  Detail & Related papers  (2025-07-11T08:00:47Z)
• ReCUT: Balancing Reasoning Length and Accuracy in LLMs via Stepwise Trails and Preference Optimization [16.51303604678232]
  Reasoning Compression ThroUgh Stepwise Trials (ReCUT) is a novel method aimed at balancing the accuracy and length of reasoning trajectory.<n> Experimental results across multiple math reasoning datasets and backbone models demonstrate that ReCUT significantly reduces reasoning lengths by approximately 30-50%.
  arXiv  Detail & Related papers  (2025-06-12T15:43:01Z)
• 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)
• Self-Refining Diffusion Samplers: Enabling Parallelization via Parareal Iterations [53.180374639531145]
  Self-Refining Diffusion Samplers (SRDS) retain sample quality and can improve latency at the cost of additional parallel compute.<n>We take inspiration from the Parareal algorithm, a popular numerical method for parallel-in-time integration of differential equations.
  arXiv  Detail & Related papers  (2024-12-11T11:08:09Z)
• DCR: Divide-and-Conquer Reasoning for Multi-choice Question Answering with LLMs [9.561022942046279]
  We propose Divide and Conquer Reasoning (DCR) to enhance the reasoning capability of large language models (LLMs) We first categorize questions into two subsets based on confidence score ($mathcalCS$), which is estimated by statistical frequency of generated answers. In particular, we first categorize questions into two subsets based on confidence score ($mathcalCS$), which is estimated by statistical frequency of generated answers.
  arXiv  Detail & Related papers  (2024-01-10T14:38:46Z)

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.