mlsanityData to 14 Oct 2026

DiMOS: Doob-Guided Inference-Time Multi-Objective Search for Scientific Design

arXiv 2610.05808 ↗First posted 5 Oct 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Scientific design often requires jointly satisfying multiple objectives and constraints. Pretrained masked diffusion models provide a generative foundation for this task, but fine-tuning them to meet these objectives and constraints incurs additional training costs, motivating inference-time guidance with frozen models. However, such guidance faces two challenges: pass-or-fail constraints and black-box reward models may provide no useful gradients, while jointly satisfying multiple requirements can leave a small feasible region, making feasible designs difficult to find within a limited inference budget. To address these challenges, we introduce DiMOS, a training-free framework for multi-objective scientific design. Using joint rewards from candidate completions, DiMOS performs approximate Doob-guided local resampling without requiring reward gradients. To allocate computation efficiently, it uses budget-efficient trajectory search to focus computation on promising continuations. Across six DNA, protein, and RNA tasks, DiMOS attains the highest joint success rate at comparable generation times, up to 1.98× the strongest baseline on DNA and protein, while maintaining high sequence uniqueness and naturalness.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 26 of its 41 references that are in the map. First posted on arXiv 5 Oct 2026.

Builds on

26 of 41 references.

References · 26

  1. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  2. 2023Training Diffusion Models with Reinforcement Learningcited by 982
  3. 2024Simple and Effective Masked Diffusion Language Modelscited by 848
  4. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 745
  5. 2024Simplified and Generalized Masked Diffusion for Discrete Datacited by 626
  6. 2018Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Reviewcited by 482
  7. 2023Directly Fine-Tuning Diffusion Models on Differentiable Rewardscited by 448
  8. 2023Practical and Asymptotically Exact Conditional Sampling in Diffusion Modelscited by 232
  9. 2025A General Framework for Inference-time Scaling and Steering of Diffusion Modelscited by 224
  10. 2023Protein Design with Guided Discrete Diffusioncited by 157
  11. 2024Simple Guidance Mechanisms for Discrete Diffusion Modelscited by 155
  12. 2024Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decodingcited by 136
14 more
  1. 2024Unlocking Guidance for Discrete State-Space Diffusion and Flow Modelscited by 129
  2. 2022Accelerating Bayesian Optimization for Biological Sequence Design with Denoising Autoencoderscited by 102
  3. 2025Inference-Time Alignment in Diffusion Models with Reward-Guided Generation: Tutorial and Reviewcited by 88
  4. 2024Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Designcited by 79
  5. 2024DEFT: Efficient Fine-Tuning of Diffusion Models by Learning the Generalised h-transformcited by 46
  6. 2024PepTune: De Novo Generation of Therapeutic Peptides with Multi-Objective-Guided Discrete Diffusioncited by 37
  7. 2024Steering Masked Discrete Diffusion Models via Discrete Denoising Posterior Predictioncited by 32
  8. 2025Reward-Guided Iterative Refinement in Diffusion Models at Test-Time with Applications to Protein and DNA Designcited by 25
  9. 2025h-Edit: Effective and Flexible Diffusion-Based Editing via Doob's h-Transformcited by 19
  10. 2025Multi-Objective-Guided Discrete Flow Matching for Controllable Biological Sequence Designcited by 15
  11. 2025Discrete Diffusion Trajectory Alignment via Stepwise Decompositioncited by 12
  12. 2024Plug-and-Play Controllable Generation for Discrete Masked Modelscited by 11
  13. 2026Training-Free Adaptation of Diffusion Models via Doob's h-Transformcited by 11
  14. 2025Ctrl-DNA: Controllable Cell-Type-Specific Regulatory DNA Design via Constrained RLcited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026VGAS: Variance-Reduced Guidance and Adaptive Selection for Training-Free Reward Alignment in Discrete Diffusion
  2. 2026InferOpt: Constrained Multi-Objective Search for LLM Inference Configurations
  3. 2026Idea Search: Guiding Tree Search with Ideas to Explore Diverse Scientific Methods
  4. 2026MORAD: Reinforcement Learning with Multi-Objective Rewards for RNA Inverse Design
  5. 2026Ligand-Conditioned Discrete Diffusion for Protein Sequence-Structure Co-Design
  6. 2026SciDesignBench: Benchmarking and Improving Language Models for Scientific Inverse Design
  7. 2026On the Design Space of Discrete Diffusion Online Adaptation for Molecular Optimization
  8. 2026RosettaSearch: Multi-Objective Inference-Time Search for Protein Sequence Design