mlsanityData to 7 Oct 2026

RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies

arXiv 2603.04639 ↗First posted 4 Mar 2026cs.RO cs.AIcited by 39

On the map: Vision-language-action models → Reasoning and chain-of-thought in VLA

Abstract (arXiv)

Memory is critical for long-horizon and history-dependent robotic manipulation. Such tasks often involve counting repeated actions or manipulating objects that become temporarily occluded. Recent vision-language-action (VLA) models have begun to incorporate memory mechanisms; however, their evaluations remain confined to narrow, non-standardized settings. This limits systematic understanding, comparison, and progress measurement. To address these challenges, we introduce RoboMME: a large-scale standardized benchmark for evaluating and advancing VLA models in long-horizon, history-dependent scenarios. Our benchmark comprises 16 manipulation tasks constructed under a carefully designed taxonomy that evaluates temporal, spatial, object, and procedural memory. We further develop a suite of 14 memory-augmented VLA variants built on the π0.5 backbone to systematically explore different memory representations across multiple integration strategies. Experimental results show that the effectiveness of memory representations is highly task-dependent, with each design offering distinct advantages and limitations across different tasks. Videos and code can be found at our website https://robomme.github.io.

Key facts. Cited by 39 papers in the map (data through 7 Oct 2026). Placed in Reasoning and chain-of-thought in VLA, one of 13 directions in Vision-language-action models. Builds on 7 of its 53 references that are in the map. First posted on arXiv 4 Mar 2026.

Builds on

7 of 53 references.

References · 7

  1. 2025SigLIP 2: Multilingual Vision-Language Encoders with Improved Semantic Understanding, Localization, and Dense Featurescited by 1,127
  2. 2020robosuite: A Modular Simulation Framework and Benchmark for Robot Learningcited by 484
  3. 2024CogACT: A Foundational Vision-Language-Action Model for Synergizing Cognition and Action in Robotic Manipulationcited by 411
  4. 2023Scaling Transformer to 1M tokens and beyond with RMTcited by 97
  5. 2025Gemini Robotics 1.5: Pushing the Frontier of Generalist Robots with Advanced Embodied Reasoning, Thinking, and Motion Transfercited by 86
  6. 2025ContextVLA: Vision-Language-Action Model with Amortized Multi-Frame Contextcited by 34
  7. 2026A Systematic Study of Data Modalities and Strategies for Co-training Large Behavior Models for Robot Manipulationcited by 20

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 39.

39 papers.

Sort:

By area:

  1. 2026World Action Models: The Next Frontier in Embodied AIcited by 46
  2. 2026EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policiescited by 18
  3. 2026World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetrycited by 13
  4. 2026RoboTTT: Context Scaling for Robot Policiescited by 13
  5. 2026World Action Models: A Surveycited by 10
  6. 2026vla-eval: A Unified Evaluation Harness for Vision-Language-Action Modelscited by 7
  7. 2026μVLA: On Recurrent Memory for Partially Observable Manipulation in VLA Modelscited by 7
  8. 2026DAM-VLA: Decoupled Asynchronous Multimodal Vision Language Action modelcited by 6
  9. 2026KEMO: Event-Driven Keyframe Memory for Long-Horizon Robot Manipulation with VLA Policiescited by 5
  10. 2026Data Pyramid for Embodied Manipulation: A Surveycited by 4
  11. 2026HarnessWAM: Bridging Prediction and Deliberation in World Action Modelscited by 4
  12. 2026Don't Drop the BATON: Long-Horizon Robot Manipulation via Agentic Subtask Exploration and Transition-aware Memorycited by 4
27 more
  1. 2026VLA-Pro: Cross-Task Procedural Memory Transfer for Vision-Language-Action Modelscited by 2
  2. 2026AURA: Action-Gated Memory for Robot Policies at Constant VRAMcited by 2
  3. 2026Training and Evaluating Diffusion Policies with Long Context Lengthscited by 2
  4. 2026WeaveLA: Event Driven Cross-Subtask Latent Memory Weaving for Repetitive Robot Manipulationcited by 2
  5. 2026MemoryVAM: Integrating Memory into Video Action Model for Robot Manipulationcited by 2
  6. 2026ChainVLA: Chaining Vision-Language-Action Queries through a Unified Execution State for Long-Horizon Manipulationcited by 2
  7. 2026SimpleMemVLA: A Simple but Effective Native-Video Memory for Vision-Language-Action Modelscited by 2
  8. 2026WatchAct: A Benchmark for Behavior-Grounded Robot Manipulationcited by 1
  9. 2026How Should Vision-Language-Action Models Use Proprioceptive State?cited by 1
  10. 20262AM: Grounding Agent-Side Memory as Guidance for Steerable Action Models in Long-Horizon Manipulationcited by 1
  11. 2026DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?cited by 0
  12. 2026TAP-VLA: Tactile Annotation Prompting for Vision Language Action Modelscited by 0
  13. 2026WorldToken: Time-First Sequence Modeling for Robotic Imitation Learningcited by 0
  14. 2026PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Controlcited by 0
  15. 2026AGM: Achievement-Grounded Memory for Closed-Loop Agents with Frozen VLA Policiescited by 0
  16. 2026MEMOBench: A Process Level Memory Benchmark for Robotic Manipulationcited by 0
  17. 2026SIMLIFE: Pattern Understanding for Long-Horizon Human-Agent Partnershipcited by 0
  18. 2026FAN: Foresight Action Normalization for Continual Adaptation of Vision-Language-Action Modelscited by 0
  19. 2026ActiveArena: Benchmarking and Understanding Active Perception in Robotic Manipulationcited by 0
  20. 2026EmbodiedMemory-Bench: Benchmarking Embodied Memory for Long-Horizon Embodied Taskscited by 0
  21. 2026RoboIRGBench: Benchmarking Implicit Referential Grounding in Vision-Language-Action Modelscited by 0
  22. 2026D2-VLA: Dual-Memory Dual-Frequency Vision-Language-Action Model For Long Dynamic Manipulationcited by 0
  23. 2026Revision, Not Restart: Revisable Visual Plans for Closed-Loop World-Action Modelscited by 0
  24. 2026Simple Agentic Memory for Generalist Robot Policiescited by 0
  25. 2026Remember What You Did: Action-History Memory with Dual-Expert Denoising for Long-Horizon Vision-Language-Action Policiescited by 0
  26. 2026Beyond the Current Scene: Event-Referential Grasping with Active View Selectioncited by 0
  27. 2026Divide-and-Remember: Recursive Action-Relevant Memory for Long-Horizon VLA Policiescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026T2Mem: Learning Test-Time Memory for Robotics ↗
  2. 2026MemoryWAM: Efficient World Action Modeling with Persistent Memory
  3. 2026RoboMemArena: A Comprehensive and Challenging Robotic Memory Benchmark
  4. 2026MIKASA-Robo-VLA: Benchmarking Memory in VLA Models for Long-Horizon Manipulation
  5. 2026ECoMEM: Explicit Concept Memory for Memory-Dependent Robot Control ↗
  6. 2025MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulation
  7. 2026RoboMME-Interference: Benchmarking Robot Memory Under Interference
  8. 2026StateMem: Single-State Residual Memory with Adaptive Inference for Vision-Language-Action Policies