Escaping Confidence Trap: Evolutionary Decoding for Mathematical Reasoning in Diffusion LLMs
Abstract
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.
Cite
@article{arxiv.2608.00605,
title = {Escaping Confidence Trap: Evolutionary Decoding for Mathematical Reasoning in Diffusion LLMs},
author = {Zhenhong Sun and Hanqing Zhao and Yatao Bian and Rongcheng Tu and Liuyue Xie and Xu Zhang and Jue Wang and Davide Modolo and Daoyi Dong and Dacheng Tao},
journal= {arXiv preprint arXiv:2608.00605},
year = {2026}
}