mlsanityData to 7 Oct 2026

FluxLite: Inference-Time Proposal Control for Discrete Diffusion Models

arXiv 2609.35947 ↗First posted 28 Sep 2026cs.LG stat.CO stat.MLcited by 1

On the map: Diffusion models → Diffusion foundations and sampling also fits: Discrete and text diffusion

Abstract (arXiv)

Many inference-time tasks for pretrained discrete diffusion models and diffusion language models reduce to drawing samples from a tilted version of the pretrained distribution. Feynman-Kac sequential Monte Carlo (SMC) makes this correction exact in principle, but its prescribed weights routinely degenerate when the proposal dynamics are misaligned with the tilt, capping the practical gains from additional particles. We introduce FluxLite, a lightweight, training-free proposal-control framework for discrete diffusion. On the sparse directed graph of pretrained reverse rates, any sparse jump-rate perturbation can be exactly compensated by a qt-weighted graph-divergence term in the Feynman-Kac potential; the target path is therefore preserved while the residual reweighting variance becomes a local convex objective. We instantiate this principle as two practical samplers: a one-hop local reallocation rule (HEU) and a small nonnegative quadratic program over pretrained-rate bases (D-VCG). We further prove population stability under the standard score-entropy training loss, identifying a tilted-path coverage factor that governs robustness to score error, together with finite-particle convergence for a fixed controlled Feynman-Kac recursion. Empirically, FluxLite improves over standard Feynman-Kac SMC baselines by up to two orders of magnitude in terminal KL on an analytically tractable finite-state CTMC benchmark, and reduces row-correlation MSE on 2D Ising sampling by 5-7x in geometric mean and up to 55x at peak.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 152 of its 190 references that are in the map. First posted on arXiv 28 Sep 2026.

Builds on

152 of 190 references.

Closest ancestors · 40

Same direction, newest first

  1. 2026Simple Approximation and Derivative Free Inference-Time Scaling for Diffusion Models via Sequential Monte Carlo on Path Measures18 May
  2. 2026Training-Free Adaptation of Diffusion Models via Doob's h-Transform18 Feb
  3. 2026Discrete Feynman-Kac Correctors15 Jan
  4. 2025Test-time scaling of diffusions with flow maps27 Nov
  5. 2025Learn to Guide Your Diffusion Model1 Oct
  6. 2025DriftLite: Lightweight Drift Control for Inference-Time Scaling of Diffusion Models25 Sep
  7. 2025MDNS: Masked Diffusion Neural Sampler via Stochastic Optimal Control14 Aug
  8. 2025Noise Hypernetworks: Amortizing Test-Time Compute in Diffusion Models13 Aug
  9. 2025Improving Classifier-Free Guidance in Masked Diffusion: Low-Dim Theoretical Insights with High-Dim Impact11 Jul
  10. 2025Diffusion Tree Sampling: Scalable inference-time alignment of diffusion models25 Jun
  11. 2025What Exactly Does Guidance Do in Masked Discrete Diffusion Models12 Jun
  12. 2025Inference-time Scaling of Diffusion Models through Classical Search29 May
28 more
  1. 2025Inference-Time Scaling of Discrete Diffusion Models via Importance Weighting and Optimal Proposal Design28 May
  2. 2025Adaptive Diffusion Guidance via Stochastic Optimal Control25 May
  3. 2025Test-Time Scaling of Diffusion Models via Noise Trajectory Search24 May
  4. 2025Diffusion Models with Double Guidance: Generate with aggregated datasets19 May
  5. 2025Preference-Guided Diffusion for Multi-Objective Offline Optimization21 Mar
  6. 2025Feynman-Kac Correctors in Diffusion: Annealing, Guidance, and Product of Experts4 Mar
  7. 2025Debiasing Guidance for Discrete Diffusion with Sequential Monte Carlo10 Feb
  8. 2025Fast Solvers for Discrete Diffusion Models: Theory and Applications of High-Order Algorithms1 Feb
  9. 2025Inference-Time Alignment in Diffusion Models with Reward-Guided Generation: Tutorial and Review16 Jan
  10. 2025A General Framework for Inference-time Scaling and Steering of Diffusion Models12 Jan
  11. 2025Test-time Alignment of Diffusion Models without Reward Over-optimization10 Jan
  12. 2024The Superposition of Diffusion Models Using the Ito Density Estimator23 Dec
  13. 2024Simple Guidance Mechanisms for Discrete Diffusion Models13 Dec
  14. 2024How Discrete and Continuous Diffusion Meet: Comprehensive Analysis of Discrete Diffusion Models via a Stochastic Integral Framework4 Oct
  15. 2024TFG: Unified Training-Free Guidance for Diffusion Models24 Sep
  16. 2024What does guidance do? A fine-grained analysis in a simple setting19 Sep
  17. 2024Classifier-Free Guidance is a Predictor-Corrector16 Aug
  18. 2024Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-Based Decoding15 Aug
  19. 2024Accelerating Diffusion Models with Parallel Sampling: Inference at Sub-Linear Time Complexity24 May
  20. 2024Theoretical Insights for Diffusion Guidance: A Case Study for Gaussian Mixture Models3 Mar
  21. 2023Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMC22 Feb
  22. 2023Optimizing DDPM Sampling with Shortcut Fine-Tuning31 Jan
  23. 2022A Continuous Time Framework for Discrete Denoising Models30 May
  24. 2021Maximum Likelihood Training of Score-Based Diffusion Models22 Jan
  25. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov
  26. 2020Denoising Diffusion Implicit Models6 Oct
  27. 2020Denoising Diffusion Probabilistic Models19 Jun
  28. 2020Improved Techniques for Training Score-Based Generative Models16 Jun

Other references · 112

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2023GPT-4 Technical Reportcited by 18,132
  4. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  5. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  6. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  7. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  8. 2020Scaling Laws for Neural Language Modelscited by 6,411
  9. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  10. 2022Flow Matching for Generative Modelingcited by 5,426
  11. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  12. 2023Let's Verify Step by Stepcited by 3,598
38 more
  1. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,550
  2. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,155
  3. 2022Training Compute-Optimal Large Language Modelscited by 2,927
  4. 2024Expanding Performance Boundaries of Open-Source Multimodal Models with Model, Data, and Test-Time Scalingcited by 1,573
  5. 2025s1: Simple test-time scalingcited by 1,311
  6. 2022MaskGIT: Masked Generative Image Transformercited by 1,082
  7. 2022Diffusion-LM Improves Controllable Text Generationcited by 1,006
  8. 2016HyperNetworkscited by 1,003
  9. 2022Building Normalizing Flows with Stochastic Interpolantscited by 964
  10. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 946
  11. 2023Stochastic Interpolants: A Unifying Framework for Flows and Diffusionscited by 897
  12. 2025Large Language Diffusion Modelscited by 872
  13. 2023Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesiscited by 847
  14. 2024Simple and Effective Masked Diffusion Language Modelscited by 800
  15. 2023Diffusion Model Alignment Using Direct Preference Optimizationcited by 767
  16. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 712
  17. 2022DiGress: Discrete Denoising diffusion for graph generationcited by 503
  18. 2024Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Modelcited by 468
  19. 2024Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Modelscited by 282
  20. 2025Mercury: Ultra-Fast Language Models Based on Diffusioncited by 225
  21. 2025A Survey on Test-Time Scaling in Large Language Models: What, How, Where, and How Well?cited by 215
  22. 2024Scaling up Masked Diffusion Models on Textcited by 195
  23. 2025d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learningcited by 178
  24. 2025Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inferencecited by 173
  25. 2023Protein Design with Guided Discrete Diffusioncited by 154
  26. 2025LLaDA2.0: Scaling Up Diffusion Language Models to 100Bcited by 154
  27. 2017MuseGAN: Multi-track Sequential Generative Adversarial Networks for Symbolic Music Generation and Accompanimentcited by 151
  28. 2024Unlocking Guidance for Discrete State-Space Diffusion and Flow Modelscited by 124
  29. 2024BoNBoN Alignment for Large Language Models and the Sweetness of Best-of-n Samplingcited by 120
  30. 2024Learning Multi-dimensional Human Preference for Text-to-Image Generationcited by 102
  31. 2025S*: Test Time Scaling for Code Generationcited by 95
  32. 2025Is Best-of-N the Best of Them? Coverage, Scaling, and Optimality in Inference-Time Alignmentcited by 91
  33. 2024Probabilistic Inference in Language Models via Twisted Sequential Monte Carlocited by 85
  34. 2024Generator Matching: Generative modeling with arbitrary Markov processescited by 82
  35. 2025Scaling Test-Time Compute Without Verification or RL is Suboptimalcited by 82
  36. 2023Sequential Monte Carlo Steering of Large Language Models using Probabilistic Programscited by 79
  37. 2025Adjoint Sampling: Highly Scalable Diffusion Samplers via Adjoint Matchingcited by 78
  38. 2024Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Designcited by 72

50 most cited of 112 All 112 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Error-Corrected Inference-Time Scaling for Imperfect Diffusion Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Discrete Diffusion Inference-Time Control with Nested Sequential Monte Carlo
  2. 2025Inference-Time Scaling of Diffusion Language Models via Trajectory Refinement
  3. 2026D-DOIT: Training-free Adaptation of Discrete Diffusion via Doob's h-Transform
  4. 2025TESS 2: A Large-Scale Generalist Diffusion Language Model
  5. 2026Consistent Diffusion Language Models
  6. 2026Optimizers for Diffusion Models: A Controlled Benchmark
  7. 2026Self-Rewarding Sequential Monte Carlo for Masked Diffusion Language Models
  8. 2025Navigating the Exploration-Exploitation Tradeoff in Inference-Time Scaling of Diffusion Models