Research map / Large language models / Language, safety and interpretability
Constrained Decoding of Diffusion LLMs with Context-Free Grammars
Niels Mündler, Jasper Dekoninck, Martin Vechev
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add CFG constrained decoding (without completion) | infilling LLMs · HumanEval C++ 1-MRI, 2-MRI, 3-MRI | syntactic correctness: +5.2%, +22.5%, +31.5% absolute | Improved |
| complete aborted instances by sampling from the intersection language (Con. vs Con.-) (ablation) | constrained decoding · HumanEval C++ MRI | syntactic correctness: 95.8% syntactically valid | Improved |
| add constrained decoding | infilling LLMs · HumanEval C++ MRI | functional correctness (pass@1): +2.4% (Con.-), +2.8% (Con.) | Improved |
| add constrained decoding (without completion) | diffusion LLMs · C++, JSON, SMILES | syntactic correctness: +16.1%, +14.7%, +26.0% | Improved |
| complete aborted instances from the intersection language (Con. vs Con.-) (ablation) | DreamCoder 7B · C++ | syntactic correctness: 19.0 -> 99.2 | Improved |
| add constrained decoding | diffusion LLMs · C++, JSON, SMILES | functional correctness: +1.9% (Con.-), +2.2% (Con.) | Improved |
| add constrained decoding | Dream 7B · JSON | functional correctness: +6.9% | Improved |
| add syntactic constraints | diffusion LLMs · SMILES | functional correctness: +0.2% average, from 1.5% baseline | No clear effect |
| increase number of infilling regions (runtime cost of constraining) | constrained decoding · HumanEval C++ MRI | runtime overhead: 67% -> 205% | Hurt |
| model size effect on constraint overhead | constrained decoding · HumanEval C++ FIM/MRI | runtime overhead: 320% (1.3B), 100% (7B), 20% (33B) | Mixed |
| add constrained decoding | diffusion LLMs · C++, JSON, SMILES | runtime: median overhead 30% | Mixed |
Closest papers by meaning
- Constrained Decoding for Diffusion Language Models via Efficient Inference over Finite Automata (NeurIPS 2026 · similarity 0.93)
- DINGO: Constrained Inference for Diffusion LLMs (NeurIPS 2025 · similarity 0.91)
- Free Draft-and-Verification: Toward Lossless Parallel Decoding for Diffusion Large Language Models (NeurIPS 2026 · similarity 0.90)
- Grammar-Aligned Decoding (NeurIPS 2024 · similarity 0.90)
- Earley-Driven Dynamic Pruning for Efficient Structured Decoding (ICML 2025 · similarity 0.89)
- Dynamically Structured Diffusion Language Model Decoding via Bayesian Inference (NeurIPS 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
