DiBS: Diffusion-Informed Branch Selection
Quick Answer
DiBS introduces a diffusion model-guided approach for Sudoku solving, enhancing branch selection in symbolic solvers.
Quick Take
It significantly reduces search costs on the Royle 17-clue benchmark, outperforming traditional heuristics in nodes and backtracks. The method ensures completeness while leveraging learned global guidance for challenging instances.
Key Points
- DiBS combines symbolic solvers with diffusion models for improved Sudoku solving.
- The approach significantly reduces search costs, especially in backtracks and long-tail percentiles.
- Experiments show DiBS outperforms strong heuristic baselines on challenging Sudoku instances.
- The method maintains completeness while providing effective global guidance.
- All related code is publicly available on GitHub.
Paper Resources
Source Excerpt
arXiv:2606. 06518v1 Announce Type: new Abstract: Sudoku is a representative constraint satisfaction problem that requires global structural reasoning under strict discrete constraints. The existing works of solving Sudoku mainly focus on two dominant approaches, i. e. , traditional heuristic and deep learning solver. However, they suffer from two complementary limitations: learning-based solvers lack hard correctness guarantees, while complete symbolic solvers are still prone to long-tail search.
To address these shortcomings, we propose a novel diffusion model-guided approach, termed as DiBS, for the branch selection search process. …
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →RAIL Guard: Closing the Evaluation-to-Remediation Gap in Responsible AI for Agents
RAIL Guard introduces a closed-loop AI pipeline for large language models (LLMs) that evaluates outputs across eight dimensions and iteratively remediates failures, achieving 96.9% convergence compared to 49.1% for traditional block-and-retry methods. The system reduces unsafe agent executions by 33% without impacting task completion and is available as open-source SDKs.