mlsanityData to 7 Oct 2026

GLaD: Geometric Latent Distillation for Vision-Language-Action Models

arXiv 2512.09619 ↗First posted 10 Dec 2025cs.ROcited by 10

On the map: LLM reasoning → Spatial and embodied reasoning

also in 1 more
Abstract (arXiv)

Most existing Vision-Language-Action (VLA) models rely primarily on RGB information, while ignoring geometric cues crucial for spatial reasoning and manipulation. In this work, we introduce GLaD, a geometry-aware VLA framework that incorporates 3D geometric priors during pretraining through knowledge distillation. Rather than distilling geometric features solely into the vision encoder, we align the LLM's hidden states corresponding to visual tokens with features from a frozen geometry-aware vision transformer (VGGT), ensuring that geometric understanding is deeply integrated into the multimodal representations that drive action prediction. Pretrained on the Bridge dataset with this geometry distillation mechanism, GLaD achieves 94.1% average success rate across four LIBERO task suites, outperforming UniVLA (92.5%) which uses identical pretraining data. These results validate that geometry-aware pretraining enhances spatial reasoning and policy generalization without requiring explicit depth sensors or 3D annotations.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning. Builds on 82 of its 85 references that are in the map. First posted on arXiv 10 Dec 2025.

Builds on

82 of 85 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Vid-LLM: A Compact Video-based 3D Multimodal LLM with Reconstruction-Reasoning Synergy29 Sep
  2. 20253DRS: MLLMs Need 3D-Aware Representation Supervision for Scene Understanding2 Jun
  3. 2025PointVLA: Injecting the 3D World into Vision-Language-Action Models10 Mar

Other references · 79

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  4. 2015Distilling the Knowledge in a Neural Networkcited by 10,076
  5. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  6. 2023Sigmoid Loss for Language Image Pre-Trainingcited by 3,048
  7. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Controlcited by 2,993
  8. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
  9. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  10. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  11. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,674
  12. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,297
38 more
  1. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  2. 2021Vision Transformers for Dense Predictioncited by 1,789
  3. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  4. 2014FitNets: Hints for Thin Deep Netscited by 1,619
  5. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,526
  6. 2025VGGT: Visual Geometry Grounded Transformercited by 1,475
  7. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  8. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  9. 2023WizardLM: Empowering large pre-trained language models to follow complex instructionscited by 1,116
  10. 2016Paying More Attention to Attention: Improving the Performance of Convolutional Neural Networks via Attention Transfercited by 1,059
  11. 2022STaR: Bootstrapping Reasoning With Reasoningcited by 895
  12. 2019Relational Knowledge Distillationcited by 873
  13. 2023Enhancing Chat Language Models by Scaling High-quality Instructional Conversationscited by 825
  14. 2023WizardCoder: Empowering Code Large Language Models with Evol-Instructcited by 728
  15. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 670
  16. 2022Large Language Models Can Self-Improvecited by 656
  17. 2024Self-Rewarding Language Modelscited by 654
  18. 2023ZoeDepth: Zero-shot Transfer by Combining Relative and Metric Depthcited by 646
  19. 2023WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instructcited by 618
  20. 2023RLAIF vs. RLHF: Scaling Reinforcement Learning from Human Feedback with AI Feedbackcited by 615
  21. 2023Textbooks Are All You Needcited by 561
  22. 2023Textbooks Are All You Need II: phi-1.5 technical reportcited by 559
  23. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 502
  24. 2023Zephyr: Direct Distillation of LM Alignmentcited by 498
  25. 2023Reinforced Self-Training (ReST) for Language Modelingcited by 486
  26. 2023MAmmoTH: Building Math Generalist Models through Hybrid Instruction Tuningcited by 483
  27. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 411
  28. 2023UltraFeedback: Boosting Language Models with Scaled AI Feedbackcited by 394
  29. 2025UniVLA: Learning to Act Anywhere with Task-centric Latent Actionscited by 386
  30. 2023Orca: Progressive Learning from Complex Explanation Traces of GPT-4cited by 369
  31. 2023LLM-QAT: Data-Free Quantization Aware Training for Large Language Modelscited by 320
  32. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 320
  33. 2023Principle-Driven Self-Alignment of Language Models from Scratch with Minimal Human Supervisioncited by 318
  34. 2023PandaLM: An Automatic Evaluation Benchmark for LLM Instruction Tuning Optimizationcited by 311
  35. 2025π3: Permutation-Equivariant Visual Geometry Learningcited by 292
  36. 2024GR-2: A Generative Video-Language-Action Model with Web-Scale Knowledge for Robot Manipulationcited by 290
  37. 2025π*0.6: a VLA That Learns From Experiencecited by 280
  38. 2024A Survey on Knowledge Distillation of Large Language Modelscited by 276

50 most cited of 79 All 79 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 10.

10 papers.

  1. 2026Geometry-Guided Camera Motion Understanding in VideoLLMscited by 2
  2. 2026VEGA: Visual Encoder Grounding Alignment for Spatially-Aware Vision-Language-Action Modelscited by 2
  3. 2026DepthCache: Depth-Guided Training-Free Visual Token Merging for Vision-Language-Action Model Inferencecited by 1
  4. 2026X-Imitator: Spatial-Aware Imitation Learning via Bidirectional Action-Pose Interactioncited by 0
  5. 2026Geometric Foundation Model Distillation for Efficient Lunar 3D Reconstructioncited by 0
  6. 2026Mind-VLA: Instruction-Aware Spatial Representation Alignment for Vision-Language-Action Modelscited by 0
  7. 2026Temporal Forcing: 4D Representation Alignment for Vision-Language-Action Modelscited by 0
  8. 2026FOCAL-VLA: Subtask-Guided Geometry Distillation and Implicit World Modeling for Vision-Language-Action Modelscited by 0
  9. 2026Spatial Grafting: Grounding 3D Features for Flow-Matching Robot Policiescited by 0
  10. 2026Where Predictive Supervision Goes Shapes What VLA Policies Learncited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026GaussVLA: Geometry-Aware Spatial Reasoning for Vision-Language-Action Model
  2. 2026StrataVLA: Hierarchical and Efficient 3D Geometric Grounding for Vision-Language-Action Models
  3. 20263DThinkVLA: Endowing Vision-Language-Action Models with Latent 3D Priors via 3D-Thinking-Guided Co-training
  4. 2025GeoAware-VLA: Implicit Geometry Aware Vision-Language-Action Model
  5. 2026GeoBridge-VLA: Geometry-Aware Residual Adaptation for Vision-Language-Action Models
  6. 2026G3VLA: Geometric inductive bias for Vision-Language-Action Models
  7. 2025Seeing Space and Motion: Enhancing Latent Actions with Geometric and Dynamic Awareness for Vision-Language-Action Models
  8. 20263DVLA: Enhancing Vision-Language-Action Models via 3D Spatial and Instance Understanding