mlsanityData to 7 Oct 2026

Dual Latent Memory for Visual Multi-agent System

arXiv 2602.00471 ↗First posted 31 Jan 2026cs.AI cs.CVcited by 9

Not on the map. Why?

Abstract (arXiv)

While Visual Multi-Agent Systems (VMAS) promise to enhance comprehensive abilities through inter-agent collaboration, empirical evidence reveals a counter-intuitive "scaling wall": increasing agent turns often degrades performance while exponentially inflating token costs. We attribute this failure to the information bottleneck inherent in text-centric communication, where converting perceptual and thinking trajectories into discrete natural language inevitably induces semantic loss. To this end, we propose Lmathbf2-VMAS, a novel model-agnostic framework that enables inter-agent collaboration with dual latent memories. Furthermore, we decouple the perception and thinking while dynamically synthesizing dual latent memories. Additionally, we introduce an entropy-driven proactive triggering that replaces passive information transmission with efficient, on-demand memory access. Extensive experiments among backbones, sizes, and multi-agent structures demonstrate that our method effectively breaks the "scaling wall" with superb scalability, improving average accuracy by 2.7-5.4% while reducing token usage by 21.3-44.8%.

Key facts. Cited by 9 papers in the map (data through 7 Oct 2026). Builds on 65 of its 78 references that are in the map. First posted on arXiv 31 Jan 2026.

Builds on

65 of 78 references.

References · 65

  1. 2017Proximal Policy Optimization Algorithmscited by 14,441
  2. 2025Qwen3-VL Technical Reportcited by 2,534
  3. 2023MMBench: Is Your Multi-modal Model an All-around Player?cited by 2,105
  4. 2023Improving Factuality and Reasoning in Language Models through Multiagent Debatecited by 1,739
  5. 2025InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiencycited by 1,419
  6. 2023MVBench: A Comprehensive Multi-modal Video Understanding Benchmarkcited by 1,086
  7. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 893
  8. 2025Why Do Multi-Agent LLM Systems Fail?cited by 559
  9. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 540
  10. 2024BLINK: Multimodal Large Language Models Can See but Not Perceivecited by 529
  11. 2024LVBench: An Extreme Long Video Understanding Benchmarkcited by 402
  12. 2023Follow Your Pose: Pose-Guided Text-to-Video Generation using Pose-Free Videoscited by 312
38 more
  1. 2025Memory in the Age of AI Agentscited by 285
  2. 2024Scaling Large Language Model-based Multi-Agent Collaborationcited by 235
  3. 2025LLaVA-OneVision-1.5: Fully Open Framework for Democratized Multimodal Trainingcited by 233
  4. 2024MuirBench: A Comprehensive Benchmark for Robust Multi-image Understandingcited by 178
  5. 2024Follow-Your-Emoji: Fine-Controllable and Expressive Freestyle Portrait Animationcited by 172
  6. 2025Towards a Science of Scaling Agent Systemscited by 125
  7. 2025Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokenscited by 120
  8. 2025Latent Visual Reasoningcited by 99
  9. 2025MemGen: Weaving Generative Latent Memory for Self-Evolving Agentscited by 99
  10. 2025Latent Collaboration in Multi-Agent Systemscited by 67
  11. 2025MDocAgent: A Multi-Modal Multi-Agent Framework for Document Understandingcited by 66
  12. 2025Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokenscited by 65
  13. 2025SimpleVQA: Multimodal Factuality Evaluation for Multimodal Large Language Modelscited by 61
  14. 2025Think with 3D: Geometric Imagination Grounded Spatial Reasoning from Limited Viewscited by 58
  15. 2025Cache-to-Cache: Direct Semantic Communication Between Large Language Modelscited by 54
  16. 2025OneThinker: All-in-one Reasoning Model for Image and Videocited by 52
  17. 2024MuMA-ToM: Multi-modal Multi-Agent Theory of Mindcited by 51
  18. 2024LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steeringcited by 50
  19. 2024SepLLM: Accelerate Large Language Models by Compressing One Segment into One Separatorcited by 41
  20. 2025Latent Sketchpad: Sketching Visual Thoughts to Elicit Multimodal Reasoning in MLLMscited by 41
  21. 2019BottleSum: Unsupervised and Self-supervised Sentence Summarization using the Information Bottleneck Principlecited by 39
  22. 2025Agentic Entropy-Balanced Policy Optimizationcited by 39
  23. 2025ASCD: Attention-Steerable Contrastive Decoding for Reducing Hallucination in MLLMcited by 36
  24. 2025MAT-Agent: Adaptive Multi-Agent Training Optimizationcited by 34
  25. 2025METAL: A Multi-Agent Framework for Chart Generation with Test-Time Scalingcited by 33
  26. 2025VisMem: Latent Vision Memory Unlocks Potential of Vision-Language Modelscited by 31
  27. 2025Visual Document Understanding and Reasoning: A Multi-Agent Collaboration Framework with Agent-Wise Adaptive Test-Time Scalingcited by 30
  28. 2025Multimodal Chain of Continuous Thought for Latent-Space Reasoning in Vision-Language Modelscited by 29
  29. 2025Thought Communication in Multiagent Collaborationcited by 28
  30. 2025Latent Chain-of-Thought for Visual Reasoningcited by 27
  31. 2025Large Language Models Miss the Multi-Agent Markcited by 25
  32. 2025Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Spacecited by 25
  33. 2024Visual Reasoning and Multi-Agent Approach in Multimodal Large Language Models (MLLMs): Solving TSP and mTSP Combinatorial Challengescited by 21
  34. 2025AdaTooler-V: Adaptive Tool-Use for Images and Videoscited by 20
  35. 2025MLP Memory: A Retriever-Pretrained Memory for Large Language Modelscited by 19
  36. 2025Auto-scaling Continuous Memory for GUI Agentcited by 19
  37. 2025Memory Decoder: A Pretrained, Plug-and-Play Memory for Large Language Modelscited by 16
  38. 2024Multi-Agent VQA: Exploring Multi-Agent Foundation Models in Zero-Shot Visual Question Answeringcited by 15

50 most cited of 65 All 65 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 9.

9 papers.

  1. 2026SEAL: Synergistic Co-Evolution of Agents and Learning Environmentscited by 9
  2. 2026The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systemscited by 8
  3. 2026SkillGraph: Self-Evolving Multi-Agent Collaboration with Multimodal Graph Topologycited by 8
  4. 2026Mem-W: Latent Memory-Native GUI Agentscited by 3
  5. 2026SPOT-E: Test-Time Entropy Shaping with Visual Spotlights for Frozen VLMscited by 1
  6. 2026Latent Action Reparameterization for Efficient Agent Inferencecited by 0
  7. 2026A Survey on Foundations and Frontiers of Multimodal Agentic Frameworks: Techniques and Applicationscited by 0
  8. 2026Collaborative Memory for Multi-Agent VLM Systemscited by 0
  9. 2026FrameMorrow: Future-guided Frame Selection with Prospective Tokens for Long-Horizon Video Generationcited by 0