mlsanityData to 14 Oct 2026

Semantic Anchoring for Robotic Action Representations

arXiv 2607.13597 ↗First posted 15 Jul 2026cs.RO cs.AI cs.CVcited by 0

On the map: Multimodal LLMs / vision-language models → Embodied and 3D vision-language models

also in 1 more
Abstract (arXiv)

Vision-Language-Action (VLA) models inherit rich semantic representations from pretrained Vision-Language Models, yet fine-tuning on limited robot demonstrations degrades this structure and undermines generalization. A fundamental question therefore arises: what constitutes a good action representation? Inspired by the mirror neuron theory's insight that observation and execution share an intention-level encoding, we examine whether a robot's action representations preserve the semantic structure captured by pretrained encoders. Systematic probing confirms that this structure erodes during finetuning, and that its quality synchronizes with both task success and out-of-distribution generalization. We further introduce a plug-and-play method that anchors action representations to a semantic manifold while decomposing representations into a shared semantic channel and a private channel, all discarded at inference, leaving the deployed model unchanged. Validated on different VLA backbones across simulation and real-world benchmarks, our method yields up to +18.7% on real-world in-distribution tasks and +21.5% on out-of-distribution generalization.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Embodied and 3D vision-language models, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 35 of its 44 references that are in the map. First posted on arXiv 15 Jul 2026.

Builds on

35 of 44 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025Don't Blind Your VLA: Aligning Visual Representations for OOD Generalization29 Oct
  2. 2025Contrastive Representation Regularization for Vision-Language-Action Models2 Oct
  3. 2025SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics2 Jun
  4. 2023RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control28 Jul

Other references · 31

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2018Representation Learning with Contrastive Predictive Codingcited by 7,777
  3. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,887
  4. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  5. 2024π0: A Vision-Language-Action Flow Model for General Robot Controlcited by 2,407
  6. 2020Shortcut Learning in Deep Neural Networkscited by 1,969
  7. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,761
  8. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,568
  9. 2024Octo: An Open-Source Generalist Robot Policycited by 1,484
  10. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  11. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,225
  12. 2022Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distributioncited by 739
19 more
  1. 2024Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Thinkcited by 703
  2. 2022R3M: A Universal Visual Representation for Robot Manipulationcited by 694
  3. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 665
  4. 2023BridgeData V2: A Dataset for Robot Learning at Scalecited by 657
  5. 2025SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Modelcited by 519
  6. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 495
  7. 2022VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Trainingcited by 395
  8. 2016Domain Separation Networkscited by 320
  9. 2026Qwen3-VL-Embedding and Qwen3-VL-Reranker: A Unified Framework for State-of-the-Art Multimodal Retrieval and Rankingcited by 230
  10. 2025LIBERO-PRO: Towards Robust and Fair Evaluation of Vision-Language-Action Models Beyond Memorizationcited by 179
  11. 2023Language-Driven Representation Learning for Roboticscited by 165
  12. 2020Better Fine-Tuning by Reducing Representational Collapsecited by 157
  13. 2024Scaling Proprioceptive-Visual Learning with Heterogeneous Pre-trained Transformerscited by 148
  14. 2025Spatial Forcing: Implicit Spatial Representation Alignment for Vision-language-action Modelcited by 134
  15. 2025FLOWER: Democratizing Generalist Robot Policies with Efficient Vision-Language-Action Flow Policiescited by 57
  16. 2024Assessing and Learning Alignment of Unimodal Vision and Language Modelscited by 32
  17. 2025Shortcut Learning in Generalist Robot Policies: The Role of Dataset Diversity and Fragmentationcited by 31
  18. 2025Modeling Fine-Grained Hand-Object Dynamics for Egocentric Video Representation Learningcited by 20
  19. 2025Embodied Representation Alignment with Mirror Neuronscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Position: Vision-Language-Action Models Cannot Be Verified to Perform Physical Reasoning
  2. 2025Enhancing Generalization in Vision-Language-Action Models by Preserving Pretrained Representations
  3. 2026Vision-Language-Action in Robotics: A Survey of Datasets, Benchmarks, and Data Engines
  4. 2026RoboSemanticBench: Diagnosing Semantic Grounding in Action Prediction for VLA Models
  5. 2026G0.5: One Autoregressive Stream for Robot Reasoning and Action
  6. 2026V-Link: Recovering Lost Visual Representations in Action DiT for Vision-Language-Action Models
  7. 2026PAIR: Bridging Perception and Action in Vision-Language-Action Models
  8. 2026Lost in Reconstruction: Aligning Action Representations with Language in Vision-Language-Action Models