Escaping Confidence Trap: Evolutionary Decoding for Mathematical Reasoning in Diffusion LLMs
On the map: LLM reasoning → Diffusion reasoning and test-time scaling
Abstract (arXiv)
Diffusion large language models (dLLMs) have emerged as a promising alternative to autoregressive LLMs, offering efficient generation through block-wise progressive unmasking. However, their strong general-purpose performance does not necessarily translate into reliable mathematical reasoning, where correctness depends on preserving coherent numerical-symbolic reasoning trajectories. In this work, we analyze the decoding trajectories of LLaDA 2.0 and identify a recurring diffusion confidence trap: local token confidence can become misaligned with global reasoning correctness during progressive block decoding. Our analysis reveals two representative failure regimes: sampling-sensitive failures, where correct paths exist but are unstable, and sampling-consistent failures, where repeated sampling converges to repetitive high-confidence but incorrect continuations. Motivated by this observation, we propose Evolutionary Decoding, a training-free test-time scaling framework that views diffusion decoding as an evolutionary process over candidate reasoning states. The framework combines step-wise selection, which preserves useful numerical-symbolic signals and suppresses repetitive patterns, with block-wise mutation, which introduces structured alternatives to escape incorrect high-confidence basins. Experiments on multiple benchmarks show that Evolutionary Decoding improves LLaDA 2.0 over confidence-based decoding, leading to more reliable mathematical reasoning.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion reasoning and test-time scaling, one of 17 directions in LLM reasoning. Builds on 32 of its 37 references that are in the map. First posted on arXiv 1 Aug 2026.
Builds on
32 of 37 references.
Closest ancestors · 11
Same direction, newest first
- 2026S3: Stratified Scaling Search for Test-Time in Diffusion Language Models7 Apr
- 2026Locally Confident, Globally Stuck: The Quality-Exploration Dilemma in Diffusion Language Models1 Apr
- 2026Adaptive Test-Time Compute Allocation for Block Diffusion Language Models in Complex Reasoning10 Feb
- 2026Prism: Efficient Test-Time Scaling via Hierarchical Search and Self-Verification for Discrete Diffusion Language Models2 Feb
- 2025Reasoning in Diffusion Large Language Models is Concentrated in Dynamic Confusion Zones19 Nov
- 2025Test-Time Scaling in Diffusion LLMs via Hidden Semi-Autoregressive Experts6 Oct
- 2025RFG: Test-Time Scaling for Diffusion Large Language Model Reasoning with Reward-Free Guidance29 Sep
- 2025Mercury: Ultra-Fast Language Models Based on Diffusion17 Jun
- 2025d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning16 Apr
- 2025Remasking Discrete Diffusion Models with Inference-Time Scaling1 Mar
- 2024Beyond Autoregression: Discrete Diffusion for Complex Reasoning and Planning18 Oct
Other references · 21
- 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
- 2021Training Verifiers to Solve Math Word Problemscited by 10,109
- 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
- 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,986
- 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
- 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,141
- 2025Large Language Diffusion Modelscited by 912
- 2024Simple and Effective Masked Diffusion Language Modelscited by 848
- 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 745
- 2024Simplified and Generalized Masked Diffusion for Discrete Datacited by 626
- 2025Dream 7B: Diffusion Large Language Modelscited by 544
- 2025Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Modelscited by 409
9 more
- 2025LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Modelscited by 285
- 2024Scaling up Masked Diffusion Models on Textcited by 199
- 2025Scalable Best-of-N Selection for Large Language Models via Self-Certaintycited by 187
- 2025LLaDA2.0: Scaling Up Diffusion Language Models to 100Bcited by 166
- 2023Likelihood-Based Diffusion Language Modelscited by 151
- 2025LLaDA-MoE: A Sparse MoE Diffusion Language Modelcited by 57
- 2026Confidence-Based Decoding is Provably Efficient for Diffusion Language Modelscited by 18
- 2025DiRL: An Efficient Post-Training Framework for Diffusion Language Modelscited by 10
- 2026Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token Spacecited by 10
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026The Confidence Shortcut: A Reasoning Failure Mode of Masked Diffusion Models
- 2026CAI-DLLM: Convergence Aware Inference for Diffusion Language Models
- 2026DreamReasoner-8B: Block-Size Curriculum Learning for Diffusion Reasoning Models
- 2026Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching
- 2026Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language Models
- 2025Beyond Surface Reasoning: Unveiling the True Long Chain-of-Thought Capacity of Diffusion Large Language Models
- 2026Preserving Mathematical Reasoning in Compressed Diffusion Language Models via Trajectory-Aware Low-Rank Approximation
- 2026UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching