Research map / Probabilistic and scientific modeling / Variational and generative sampling
Constrained Adaptive Rejection Sampling
Paweł Parys, Sairam Vaidya, Taylor Berg-Kirkpatrick, Loris D'Antoni
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 |
|---|---|---|---|
| use CARS instead of ARS (prune all invalid continuations of prefixes) (ablation) | CARS · SMILES molecular generation | samples to get 100 valid molecules: 220 -> 183 | Improved |
| use CARS instead of plain RS | CARS · SMILES molecular generation | samples to get 100 valid molecules: 793 -> 183 | Improved |
| prune beyond first token (CARS vs RSFT) (ablation) | CARS · SMILES molecular generation | samples to get 100 valid molecules: 765 -> 183 | Improved |
| use CARS instead of GCD | CARS · SMILES molecular generation | validity: 0.70 -> 0.87 | Improved |
| use CARS instead of GCD | CARS · Spider dev text-to-SQL | execution accuracy: 0.525 -> 0.578 | Improved |
| use CARS instead of ARS (ablation) | CARS · Spider dev text-to-SQL | total samples: 1435 -> 1146 | Improved |
| use CARS instead of RS | CARS · Spider dev text-to-SQL | total samples: 2126 -> 1146 | Improved |
| use CARS instead of MCMC/AWRS | CARS · Spider dev text-to-SQL | total samples: 10340 -> 1146 | Improved |
| use CARS instead of ARS (ablation) | CARS · Spider dev text-to-SQL | execution accuracy: 0.574 -> 0.578 | No clear effect |
| use CARS instead of RSFT (ablation) | CARS · XML fuzzing no grammar | generations for 100 valid samples: 893 -> 440 | Improved |
| use CARS instead of RSFT (ablation) | CARS · JSON fuzzing no grammar | generations for 100 valid samples: 601 -> 130 | Improved |
| use CARS instead of GCD | CARS · XML fuzzing no grammar | line coverage: 7,115 -> 7,985 | Improved |
Closest papers by meaning
- Constrained Sampling for Language Models Should Be Easy: An MCMC Perspective (NeurIPS 2025 · similarity 0.93)
- Constrained Decoding for Diffusion Language Models via Efficient Inference over Finite Automata (NeurIPS 2026 · similarity 0.91)
- Grammar-Aligned Decoding (NeurIPS 2024 · similarity 0.90)
- Learning How Hard to Think: Input-Adaptive Allocation of LM Computation (ICLR 2025 · similarity 0.88)
- Towards Optimal Multi-draft Speculative Decoding (ICLR 2025 · similarity 0.88)
- On the Query Complexity of Verifier-Assisted Language Generation (ICML 2025 · similarity 0.88)
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.
