Constrained Code Generation with Discrete Diffusion
1d ago
Quick Take
Constrained Diffusion for Code enhances discrete diffusion models by integrating constraint satisfaction in code generation.
Key Points
- Introduces CDC, a training-free neurosymbolic framework.
- Enhances denoising with constraint-aware operators.
- Outperforms existing models in correctness and security.
📖 Reader Mode
~2 min readAbstract:Discrete diffusion models are a powerful, emerging paradigm for code generation. They construct programs through iterative refinement of partially corrupted token sequences and enable parallel token refinement. Importantly, this paradigm exposes a global program state at each denoising step, which provides a natural intervention point for enforcing program-level functionality and security constraints, guiding the generation before the final code is committed. Building on this observation, the paper introduces Constrained Diffusion for Code (CDC), a training-free neurosymbolic inference framework that integrates constraint satisfaction directly into the reverse denoising process. CDC augments the base discrete diffusion sampler with constraint-aware denoising operators that combine mathematical optimization with program analysis to identify constraint-relevant regions of the intermediate program state and locally adjust the denoising trajectory, steering generation toward feasible programs while remaining close to the base model. Across code generation benchmarks, CDC consistently improves constraint satisfaction in functional correctness, security, and even syntax, outperforming discrete diffusion and autoregressive baselines with less corrective computation and more localized edits.
| Subjects: | Computation and Language (cs.CL); Programming Languages (cs.PL) |
| Cite as: | arXiv:2605.16829 [cs.CL] |
| (or arXiv:2605.16829v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.16829 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Lize Shao [view email]
[v1]
Sat, 16 May 2026 06:15:47 UTC (983 KB)
— Originally published at arxiv.org
More from arXiv cs.CL
See more →arXiv cs.CL·Leyao Wang, Yanan He, Peng Chen, Asaf Yehudai, Yixin Liu, Rex Ying, Michal Shmueli-Scheuer, Arman Cohan
17h ago FeaturedOriginal
Time to REFLECT: Can We Trust LLM Judges for Evidence-based Research Agents?
AI Summary
The reliability of LLM judges for evaluating deep research agents is critically assessed using the REFLECT benchmark.