mlsanityData to 7 Oct 2026

VLM-Guided Group Preference Alignment for Diffusion-based Human Mesh Recovery

arXiv 2602.19180 ↗First posted 22 Feb 2026cs.CVcited by 2

On the map: LLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment

Abstract (arXiv)

Human mesh recovery (HMR) from a single RGB image is inherently ambiguous, as multiple 3D poses can correspond to the same 2D observation. Recent diffusion-based methods tackle this by generating various hypotheses, but often sacrifice accuracy. They yield predictions that are either physically implausible or drift from the input image, especially under occlusion or in cluttered, in-the-wild scenes. To address this, we introduce a dual-memory augmented HMR critique agent with self-reflection to produce context-aware quality scores for predicted meshes. These scores distill fine-grained cues about 3D human motion structure, physical feasibility, and alignment with the input image. We use these scores to build a group-wise HMR preference dataset. Leveraging this dataset, we propose a group preference alignment framework for finetuning diffusion-based HMR models. This process injects the rich preference signals into the model, guiding it to generate more physically plausible and image-consistent human meshes. Extensive experiments demonstrate that our method achieves superior performance compared to state-of-the-art approaches.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Diffusion and multimodal model alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 58 of its 69 references that are in the map. First posted on arXiv 22 Feb 2026.

Builds on

58 of 69 references.

Closest ancestors · 6

Same direction, newest first

  1. 2025DanceGRPO: Unleashing GRPO on Visual Generation12 May
  2. 2024Aesthetic Post-Training Diffusion Models from Generic Preferences with Step-by-step Preference Optimization6 Jun
  3. 2024Aligning Diffusion Models by Optimizing Human Utility6 Apr
  4. 2023Using Human Feedback to Fine-tune Diffusion Models without Any Reward Model22 Nov
  5. 2023Diffusion Model Alignment Using Direct Preference Optimization21 Nov
  6. 2023Directly Fine-Tuning Diffusion Models on Differentiable Rewards29 Sep

Other references · 52

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  4. 2020Denoising Diffusion Implicit Modelscited by 9,655
  5. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  6. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  7. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  8. 2017Deep reinforcement learning from human preferencescited by 3,846
  9. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Stepscited by 1,961
  10. 2019Expressive Body Capture: 3D Hands, Face, and Body from a Single Imagecited by 1,600
  11. 2023Training Diffusion Models with Reinforcement Learningcited by 939
  12. 2017End-to-end Recovery of Human Shape and Posecited by 883
38 more
  1. 2023AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedbackcited by 738
  2. 2023ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debatecited by 724
  3. 2016Keep it SMPL: Automatic Estimation of 3D Human Pose and Shape from a Single Imagecited by 692
  4. 2025Flow-GRPO: Training Flow Matching Models via Online RLcited by 605
  5. 2023Prometheus: Inducing Fine-grained Evaluation Capability in Language Modelscited by 558
  6. 2016Monocular 3D Human Pose Estimation In The Wild Using Improved CNN Supervisioncited by 493
  7. 2024From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipelinecited by 484
  8. 2024From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judgecited by 480
  9. 2023DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Modelscited by 431
  10. 2023JudgeLM: Fine-tuned Large Language Models are Scalable Judgescited by 351
  11. 2023Humans in 4D: Reconstructing and Tracking Humans with Transformerscited by 344
  12. 2018Learning 3D Human Dynamics from Videocited by 315
  13. 2020HybrIK: A Hybrid Analytical-Neural Inverse Kinematics Solution for 3D Human Pose and Shape Estimationcited by 304
  14. 2023Large Language Models are Effective Text Rankers with Pairwise Ranking Promptingcited by 302
  15. 2024Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Modelscited by 301
  16. 2021PyMAF: 3D Human Pose and Shape Regression with Pyramidal Mesh Alignment Feedback Loopcited by 250
  17. 2023BEDLAM: A Synthetic Dataset of Bodies Exhibiting Detailed Lifelike Animated Motioncited by 229
  18. 2017Unite the People: Closing the Loop Between 3D and 2D Human Representationscited by 205
  19. 2021Probabilistic Modeling for Human Mesh Recoverycited by 152
  20. 2024Who Validates the Validators? Aligning LLM-Assisted Evaluation of LLM Outputs with Human Preferencescited by 142
  21. 2021SPEC: Seeing People in the Wild with an Estimated Cameracited by 134
  22. 2023DNA-Rendering: A Diverse Neural Actor Repository for High-Fidelity Human-centric Renderingcited by 103
  23. 2024TokenHMR: Advancing Human Mesh Recovery with a Tokenized Pose Representationcited by 93
  24. 2024An Empirical Study of LLM-as-a-Judge for LLM Evaluation: Fine-tuned Judge Model is not a General Substitute for GPT-4cited by 90
  25. 2023Hi4D: 4D Instance Segmentation of Close Human Interactioncited by 87
  26. 2025PromptHMR: Promptable Human Mesh Recoverycited by 70
  27. 2024CameraHMR: Aligning People with Perspectivecited by 64
  28. 2021Playing for 3D Human Recoverycited by 63
  29. 2024Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimationcited by 57
  30. 20203D Multi-bodies: Fitting Sets of Plausible 3D Human Models to Ambiguous Image Datacited by 50
  31. 2021Hierarchical Kinematic Probability Distributions for 3D Human Shape and Pose Estimation from Images in the Wildcited by 48
  32. 2023Zolly: Zoom Focal Length Correctly for Perspective-Distorted Human Mesh Reconstructioncited by 41
  33. 2024Score-Guided Diffusion for 3D Human Recoverycited by 39
  34. 2023POTTER: Pooling Attention Transformer for Efficient Human Mesh Recoverycited by 31
  35. 2023Distribution-Aligned Diffusion for Human Mesh Recoverycited by 27
  36. 2023HuManiFlow: Ancestor-Conditioned Normalising Flows on SO(3) Manifolds for Human Pose and Shape Distribution Estimationcited by 23
  37. 2025Sentient Agent as a Judge: Evaluating Higher-Order Social Cognition in Large Language Modelscited by 21
  38. 2025TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Themcited by 20

50 most cited of 52 All 52 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026MotionGRPO: Overcoming Low Intra-Group Diversity in GRPO-Based Egocentric Motion Recoverycited by 0
  2. 2026MHE-Former: Multi-Hypothesis Transformers via Entropy Maximization for 3D Mesh Recoverycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023Generative Approach for Probabilistic Human Mesh Recovery using Diffusion Models
  2. 2025Adapting Human Mesh Recovery with Vision-Language Feedback
  3. 2026DiffProxy: Multi-View Human Mesh Recovery via Diffusion-Generated Dense Proxies
  4. 2023DiffMesh: A Motion-aware Diffusion Framework for Human Mesh Recovery from Videos
  5. 2025LieHMR: Autoregressive Human Mesh Recovery with SO(3) Diffusion
  6. 2025FastHMR: Accelerating Human Mesh Recovery via Token and Layer Merging with Diffusion Decoding
  7. 2026Discriminative-Generative Synergy for Occlusion Robust 3D Human Mesh Recovery
  8. 2026PoseDreamer: Scalable and Photorealistic Human Data Generation Pipeline with Diffusion Models