mlsanityData to 7 Oct 2026

Training Strategies for Efficient Embodied Reasoning

arXiv 2505.08243 ↗First posted 13 May 2025cs.ROcited by 53

On the map: LLM reasoning → Spatial and embodied reasoning

also in 1 more
Abstract (arXiv)

Robot chain-of-thought reasoning (CoT) -- wherein a model predicts helpful intermediate representations before choosing actions -- provides an effective method for improving the generalization and performance of robot policies, especially vision-language-action models (VLAs). While such approaches have been shown to improve performance and generalization, they suffer from core limitations, like needing specialized robot reasoning data and slow inference speeds. To design new robot reasoning approaches that address these issues, a more complete characterization of why reasoning helps policy performance is critical. We hypothesize several mechanisms by which robot reasoning improves policies -- (1) better representation learning, (2) improved learning curricularization, and (3) increased expressivity -- then devise simple variants of robot CoT reasoning to isolate and test each one. We find that learning to generate reasonings does lead to better VLA representations, while attending to the reasonings aids in actually leveraging these features for improved action prediction. Our results provide us with a better understanding of why CoT reasoning helps VLAs, which we use to introduce two simple and lightweight alternative recipes for robot reasoning. Our proposed approaches achieve significant performance gains over non-reasoning policies, state-of-the-art results on the LIBERO-90 benchmark, and a 3x inference speedup compared to standard robot reasoning.

Key facts. Cited by 53 papers in the map (data through 7 Oct 2026). Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning. Builds on 54 of its 64 references that are in the map. First posted on arXiv 13 May 2025.

Builds on

54 of 64 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models27 Mar
  2. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success27 Feb
  3. 2024Robotic Control via Embodied Chain-of-Thought Reasoning11 Jul
  4. 2024SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities22 Jan

Other references · 50

  1. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  2. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  3. 2022RT-1: Robotics Transformer for Real-World Control at Scalecited by 2,007
  4. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  5. 2020Don't Stop Pretraining: Adapt Language Models to Domains and Taskscited by 1,548
  6. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  7. 2024Octo: An Open-Source Generalist Robot Policycited by 1,427
  8. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,338
  9. 2022Code as Policies: Language Model Programs for Embodied Controlcited by 1,172
  10. 2025GR00T N1: An Open Foundation Model for Generalist Humanoid Robotscited by 1,167
  11. 2024DROID: A Large-Scale In-The-Wild Robot Manipulation Datasetcited by 801
  12. 2022Fine-Tuning can Distort Pretrained Features and Underperform Out-of-Distributioncited by 722
38 more
  1. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 670
  2. 2025FAST: Efficient Action Tokenization for Vision-Language-Action Modelscited by 643
  3. 2023BridgeData V2: A Dataset for Robot Learning at Scalecited by 634
  4. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 411
  5. 2024Universal Manipulation Interface: In-The-Wild Robot Teaching Without In-The-Wild Robotscited by 373
  6. 2024TinyVLA: Towards Fast, Data-Efficient Vision-Language-Action Models for Robotic Manipulationcited by 331
  7. 20243D-VLA: A 3D Vision-Language-Action Generative World Modelcited by 320
  8. 2024TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policiescited by 298
  9. 2023Towards Revealing the Mystery behind Chain of Thought: A Theoretical Perspectivecited by 293
  10. 2024To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoningcited by 257
  11. 2024Chain of Thought Empowers Transformers to Solve Inherently Serial Problemscited by 256
  12. 2023Think before you speak: Training Language Models With Pause Tokenscited by 254
  13. 2021Bridge Data: Boosting Generalization of Robotic Skills with Cross-Domain Datasetscited by 248
  14. 2023Where are we in the search for an Artificial Visual Cortex for Embodied Intelligence?cited by 215
  15. 2022The Parallelism Tradeoff: Limitations of Log-Precision Transformerscited by 191
  16. 2025DexVLA: Vision-Language Model with Plug-In Diffusion Expert for General Robot Controlcited by 191
  17. 2024Let's Think Dot by Dot: Hidden Computation in Transformer Language Modelscited by 186
  18. 2025HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Modelcited by 176
  19. 2024Data Scaling Laws in Imitation Learning for Robotic Manipulationcited by 144
  20. 2023RT-Trajectory: Robotic Task Generalization via Hindsight Trajectory Sketchescited by 131
  21. 2024Scaling Cross-Embodied Learning: One Policy for Manipulation, Navigation, Locomotion and Aviationcited by 108
  22. 2022Grounding Language with Visual Affordances over Unstructured Datacited by 101
  23. 2023On Bringing Robots Homecited by 101
  24. 2024LLARVA: Vision-Action Instruction Tuning Enhances Robot Learningcited by 87
  25. 2022Latent Plans for Task-Agnostic Offline Reinforcement Learningcited by 81
  26. 2025OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extractioncited by 54
  27. 2024QueST: Self-Supervised Skill Abstractions for Learning Continuous Controlcited by 51
  28. 2025A0: An Affordance-Aware Hierarchical Model for General Robotic Manipulationcited by 44
  29. 2024Robot Utility Models: General Policies for Zero-Shot Deployment in New Environmentscited by 43
  30. 2024From Multimodal LLMs to Generalist Embodied Agents: Methods and Lessonscited by 42
  31. 2024Vision-Language Models Provide Promptable Representations for Reinforcement Learningcited by 41
  32. 2025Beyond Sight: Finetuning Generalist Robot Policies with Heterogeneous Sensors via Language Groundingcited by 40
  33. 2023What Makes Pre-Trained Visual Representations Successful for Robust Manipulation?cited by 38
  34. 2022Affordance Learning from Play for Sample-Efficient Policy Learningcited by 27
  35. 2022Co-training Improves Prompt-based Learning for Large Language Modelscited by 24
  36. 2024What Do Learning Dynamics Reveal About Generalization in LLM Reasoning?cited by 22
  37. 2023How to prepare your task head for finetuningcited by 18
  38. 2024Mixing It Up: The Cocktail Effect of Multi-Task Fine-Tuning on LLM Performance -- A Case Study in Financecited by 15

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 24; 2026: 29.

53 papers.

Sort:

By area:

  1. 2025SimpleVLA-RL: Scaling VLA Training via Reinforcement Learningcited by 141
  2. 2025mimic-video: Video-Action Models for Generalizable Robot Control Beyond VLAscited by 133
  3. 2025OneTwoVLA: A Unified Vision-Language-Action Model with Adaptive Reasoningcited by 105
  4. 2025Vision-Language-Action Models for Robotics: A Review Towards Real-World Applicationscited by 82
  5. 2025Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Surveycited by 66
  6. 2026Qwen-RobotManip Technical Report: Alignment Unlocks Scale for Robotic Manipulation Foundation Modelscited by 47
  7. 2025Pure Vision Language Action (VLA) Models: A Comprehensive Surveycited by 46
  8. 2025Don't Blind Your VLA: Aligning Visual Representations for OOD Generalizationcited by 34
  9. 2025Do What You Say: Steering Vision-Language-Action Models via Runtime Reasoning-Action Alignment Verificationcited by 25
  10. 2025Efficient Vision-Language-Action Models for Embodied Manipulation: A Systematic Surveycited by 25
  11. 2026Xiaomi-Robotics-1: Scaling Vision-Language-Action Models with over 100K Hours of Real-World Trajectoriescited by 24
  12. 2026Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planningcited by 20
38 more
  1. 2026A Systematic Study of Data Modalities and Strategies for Co-training Large Behavior Models for Robot Manipulationcited by 20
  2. 2025GraphCoT-VLA: A 3D Spatial-Aware Reasoning Vision-Language-Action Model for Robotic Manipulation with Ambiguous Instructionscited by 19
  3. 2026Latent Reasoning VLA: Latent Thinking and Prediction for Vision-Language-Action Modelscited by 19
  4. 2025Parallels Between VLA Model Post-Training and Human Motor Learning: Progress, Challenges, and Trendscited by 17
  5. 2025DeepThinkVLA: Enhancing Reasoning Capability of Vision-Language-Action Modelscited by 17
  6. 2026VP-VLA: Visual Prompting as an Interface for Vision-Language-Action Modelscited by 17
  7. 2025Thinking in 360deg: Humanoid Visual Search in the Wildcited by 14
  8. 2026CycleVLA: Proactive Self-Correcting Vision-Language-Action Models via Subtask Backtracking and Minimum Bayes Risk Decodingcited by 13
  9. 2025Mantis: A Versatile Vision-Language-Action Model with Disentangled Visual Foresightcited by 12
  10. 2025Unifying Perception and Action: A Hybrid-Modality Pipeline with Implicit Visual Chain-of-Thought for Robotic Action Generationcited by 12
  11. 2025PEEK: Guiding and Minimal Image Representations for Zero-Shot Generalization of Robot Manipulation Policiescited by 11
  12. 2025DualVLA: Building a Generalizable Embodied Agent via Partial Decoupling of Reasoning and Actioncited by 11
  13. 2026SteerVLA: Steering Vision-Language-Action Models in Long-Tail Driving Scenarioscited by 11
  14. 2026Flex-π: A Multi-Stream World-Action Model with Compute Flexibilitycited by 9
  15. 2025Maestro: Orchestrating Robotics Modules with Vision-Language Models for Zero-Shot Generalist Robotscited by 8
  16. 2025Guardian: Detecting Robotic Planning and Execution Errors with Vision-Language Modelscited by 8
  17. 2025Survey of Vision-Language-Action Models for Embodied Manipulationcited by 7
  18. 2025Unified Embodied VLM Reasoning with Robotic Action via Autoregressive Discretized Pre-trainingcited by 7
  19. 2026PoseVLA: Universal Pose Pretraining for Generalizable Vision-Language-Action Policiescited by 7
  20. 2025Learning Affordances at Inference-Time for Vision-Language-Action Modelscited by 6
  21. 2026Chain-of-Goals Hierarchical Policy for Long-Horizon Offline Goal-Conditioned RLcited by 5
  22. 2026TRAP: Hijacking VLA CoT-Reasoning via Adversarial Patchescited by 5
  23. 2026Training Vision-Language-Action Models with Dense Embodied Chain-of-Thought Supervisioncited by 5
  24. 2025An Anatomy of Vision-Language-Action Models: From Modules to Milestones and Challengescited by 4
  25. 2026DextER: Language-driven Dexterous Grasp Generation with Embodied Reasoningcited by 4
  26. 2026Revisiting Embodied Chain-of-Thought for Generalizable Robot Manipulationcited by 4
  27. 2026Self-Supervised Bootstrapping of Action-Predictive Embodied Reasoningcited by 3
  28. 2026RoboAlign: Learning Test-Time Reasoning for Language-Action Alignment in Vision-Language-Action Modelscited by 3
  29. 2025EVE: A Generator-Verifier System for Generative Policiescited by 2
  30. 2026SPARC: Reliable Spatial Annotations from Robot Demonstrations at Scalecited by 1
  31. 2026Do Vision-Language-Action Models Mean What They Say? On the Role of Faithfulness in Embodied Reasoningcited by 1
  32. 2026Cross-Embodiment Transfer via Behavior-Aligned Representationscited by 1
  33. 2026Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-usecited by 1
  34. 2026Environmental Understanding Vision-Language Model for Embodied Agentcited by 0
  35. 2026Beyond Thinking: Imagining in 360^circ for Humanoid Visual Searchcited by 0
  36. 2026DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?cited by 0
  37. 2026APIVOT: Adaptive Planning with Interleaved Vision-Language Thoughtscited by 0
  38. 2026Latent Memory Palace: Reasoning for Control as Autoregressive Variational Inferencecited by 0

50 most cited of 53 All 53 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026VLA-Thinker: Boosting Vision-Language-Action Models through Thinking-with-Image Reasoning
  2. 2026Steerable Vision-Language-Action Policies for Embodied Reasoning and Hierarchical Control
  3. 2025Robot-R1: Reinforcement Learning for Enhanced Embodied Reasoning in Robotics
  4. 2025Mind to Hand: Purposeful Robotic Control via Embodied Reasoning
  5. 2026Toward Unified Robot Learning: Bridging Representation, Vision-Language-Action, and World Models
  6. 2025Hybrid Training for Vision-Language-Action Models
  7. 2026G0.5: One Autoregressive Stream for Robot Reasoning and Action
  8. 2026MobileVLA-R1 2.0: RL-Enhanced Reasoning for Mobile Robot Control