AI TrackerFrontier desk · AEST
Back to news
normalaitech

Binary Discrete Diffusion: Candidate-Set Diffusion for Logic and Planning Tasks

arXiv cs.CL · 21 Oct 2026, 03:11 AEST

Binary Discrete Diffusion introduces a new approach to handling uncertainty in categorical data. This method shows significant improvements in various reasoning and planning tasks compared to traditional models.

Key facts

Read aloud

Binary Discrete Diffusion introduces a new approach to handling uncertainty in categorical data. This method shows significant improvements in various reasoning and planning tasks compared to traditional models. Discrete diffusion models iteratively denoise a noisy sequence of categorical data such as text, enabling any-order generation with self-correction and in-filling capabilities. We close this gap with Binary Discrete Diffusion, a family of continuous-time discrete-diffusion processes whose noisy state is a set of candidate tokens per position, represented as a binary membership vector. On Sudoku-Extreme, Boxoban, and graph coloring, the symmetric formulation of binary diffusion consistently outperforms strong, well-tuned masked and uniform baselines under matched architectures and controlled training, sampling, and tuning budgets. These results support candidate sets as an effective intermediate representation for iterative reasoning and show that the corruption kernel can materially change the computation learned by the reverse model.

Sources in this cluster