mlsanityData to 7 Oct 2026

WorldMM: Dynamic Multimodal Memory Agent for Long Video Reasoning

arXiv 2512.02425 ↗First posted 2 Dec 2025cs.CV cs.AI cs.CLcited by 50

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Recent advances in video large language models have demonstrated strong capabilities in understanding short clips. However, scaling them to hours- or days-long videos remains highly challenging due to limited context capacity and the loss of critical visual details during abstraction. Existing memory-augmented methods mitigate this by leveraging textual summaries of video segments, yet they heavily rely on text and fail to utilize visual evidence when reasoning over complex scenes. Moreover, retrieving from fixed temporal scales further limits their flexibility in capturing events that span variable durations. To address this, we introduce WorldMM, a novel multimodal memory agent that constructs and retrieves from multiple complementary memories, encompassing both textual and visual representations. WorldMM comprises three types of memory: episodic memory indexes factual events across multiple temporal scales, semantic memory continuously updates high-level conceptual knowledge, and visual memory preserves detailed information about scenes. During inference, an adaptive retrieval agent iteratively selects the most relevant memory source and leverages multiple temporal granularities based on the query, continuing until it determines that sufficient information has been gathered. WorldMM significantly outperforms existing baselines across five long video question-answering benchmarks, achieving an average 8.4% performance gain over previous state-of-the-art methods, showing its effectiveness on long video reasoning.

Key facts. Cited by 50 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 40 of its 49 references that are in the map. First posted on arXiv 2 Dec 2025.

Builds on

40 of 49 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025Conan: Progressive Learning to Reason Like a Detective over Multi-Scale Visual Evidence23 Oct
  2. 2025Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding15 Oct
  3. 2025ReWatch-R1: Boosting Complex Video Reasoning in Large Vision-Language Models through Agentic Data Synthesis28 Sep
  4. 2025Video-MTR: Reinforced Multi-Turn Reasoning for Long Video Understanding28 Aug
  5. 2025E-VRAG: Enhancing Long Video Understanding with Resource-Efficient Retrieval Augmented Generation3 Aug
  6. 2025Video-RTS: Rethinking Reinforcement Learning and Test-Time Scaling for Efficient and Enhanced Video Reasoning9 Jul
  7. 2024VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videos29 May

Other references · 33

  1. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  2. 2025Qwen3-VL Technical Reportcited by 2,534
  3. 2024Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysiscited by 1,375
  4. 2025Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Modelscited by 1,175
  5. 2024LLaVA-Video: Video Instruction Tuning With Synthetic Datacited by 550
  6. 2024Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexitycited by 437
  7. 2024LVBench: An Extreme Long Video Understanding Benchmarkcited by 402
  8. 2024LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understandingcited by 341
  9. 2024InternVideo2: Scaling Foundation Models for Multimodal Video Understandingcited by 310
  10. 2024LightRAG: Simple and Fast Retrieval-Augmented Generationcited by 310
  11. 2024VideoAgent: A Memory-augmented Multimodal Agent for Video Understandingcited by 224
  12. 2024MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understandingcited by 223
21 more
  1. 2025From RAG to Memory: Non-Parametric Continual Learning for Large Language Modelscited by 223
  2. 2024VideoChat-Flash: Hierarchical Compression for Long-Context Video Modelingcited by 212
  3. 2025Adaptive Keyframe Sampling for Long Video Understandingcited by 153
  4. 2025Time-R1: Post-Training Large Vision Language Model for Temporal Video Groundingcited by 138
  5. 2024Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehensioncited by 134
  6. 2024HourVideo: 1-Hour Video-Language Understandingcited by 110
  7. 2025VLM2Vec-V2: Advancing Multimodal Embedding for Videos, Images, and Visual Documentscited by 110
  8. 2024LongLLaVA: Scaling Multi-modal LLMs to 1000 Images Efficiently via a Hybrid Architecturecited by 100
  9. 2025Seeing, Listening, Remembering, and Reasoning: A Multimodal Agent with Long-Term Memorycited by 88
  10. 2023Distil-Whisper: Robust Knowledge Distillation via Large-Scale Pseudo Labellingcited by 87
  11. 2025VideoRAG: Retrieval-Augmented Generation with Extreme Long-Context Videoscited by 76
  12. 2024Too Many Frames, Not All Useful: Efficient Strategies for Long-Form Video QAcited by 68
  13. 2025Ego-R1: Chain-of-Tool-Thought for Ultra-Long Egocentric Video Reasoningcited by 67
  14. 2025VideoRAG: Retrieval-Augmented Generation over Video Corpuscited by 55
  15. 2024InternLM-XComposer2.5-OmniLive: A Comprehensive Multimodal System for Long-term Streaming Video and Audio Interactionscited by 48
  16. 2025HippoMM: Hippocampal-inspired Multimodal Memory for Long Audiovisual Event Understandingcited by 19
  17. 2025UniversalRAG: Retrieval-Augmented Generation over Corpora of Diverse Modalities and Granularitiescited by 18
  18. 2022EVEREST: Efficient Masked Video Autoencoder by Removing Redundant Spatiotemporal Tokenscited by 15
  19. 2025∞-Video: A Training-Free Approach to Long Video Understanding via Continuous-Time Memory Consolidationcited by 14
  20. 2025AdaVideoRAG: Omni-Contextual Adaptive Retrieval-Augmented Efficient Long Video Understandingcited by 14
  21. 2025Movie Facts and Fibs (MF2): A Benchmark for Long Movie Understandingcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 48.

50 papers.

Sort:

By area:

  1. 2025Memory in the Age of AI Agentscited by 285
  2. 2026A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agentscited by 40
  3. 2026From Storage to Experience: A Survey on the Evolution of LLM Agent Memory Mechanismscited by 19
  4. 2026Governing Evolving Memory in LLM Agents: Risks, Mechanisms, and the Stability and Safety Governed Memory (SSGM) Frameworkcited by 16
  5. 2025AI Meets Brain: Memory Systems from Cognitive Neuroscience to Autonomous Agentscited by 12
  6. 2026MementoGUI: Learning Agentic Multimodal Memory Control for Long-Horizon GUI Agentscited by 9
  7. 2026Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Applicationcited by 8
  8. 2026MemDreamer: Decoupling Perception and Reasoning for Long Video Understanding via Hierarchical Graph Memory and Agentic Retrieval Mechanismcited by 6
  9. 2026MA-EgoQA: Question Answering over Egocentric Videos from Multiple Embodied Agentscited by 5
  10. 2026Video-Oasis: Rethinking Evaluation of Video Understandingcited by 3
  11. 2026Memory Transfer Learning: How Memories are Transferred Across Domains in Coding Agentscited by 3
  12. 2026Bridging Modalities, Spanning Time: Structured Memory for Ultra-Long Agentic Video Reasoningcited by 3
38 more
  1. 2026FOLIO: Focused Semantic Memory for Streaming Video Understandingcited by 3
  2. 2026VideoHarness-RSI: Recursive Harness Self-Improvement for Long-Video Understanding with Frozen Vision-Language Modelscited by 3
  3. 2026Are Video Reasoning Models Ready to Go Outside?cited by 2
  4. 2026Visual Agentic Memory: Enabling Online Long Video Understanding via Online Indexing, Hierarchical Memory, and Agentic Retrievalcited by 2
  5. 2026LightMem-Ego: Your AI Memory for Everyday Lifecited by 2
  6. 2026Vinci2: Providing Proactive Assistance in Continuous Egocentric Videoscited by 2
  7. 2026MetaVideoAgent: Automated Video-Agent Evolution for Long-Form Video Understandingcited by 2
  8. 2026ShallowStream: Index Shallow then Answer Deep for Streaming Video Understandingcited by 2
  9. 2026VideoAtlas: Navigating Long-Form Video in Logarithmic Computecited by 1
  10. 2026Maestro: Reinforcement Learning to Orchestrate Hierarchical Model-Skill Ensemblescited by 1
  11. 2026Task-Focused Memorization for Multimodal Agentscited by 1
  12. 2026Rethinking RAG in Long Videos: What to Retrieve and How to Use It?cited by 1
  13. 2026HiMe: Hierarchical Embodied Memory for Long-Horizon Vision-Language-Action Controlcited by 1
  14. 2026Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and Editingcited by 1
  15. 2026Do Agents Dream of False Memories? Black-box Visual Attacks on Long-term Memory in Multimodal AI Agentscited by 1
  16. 2026Keep It Simple: Multi-Key Episodic Memory Retrieval for Ultra-Long Video Understandingcited by 1
  17. 2026REVEAL: A Rubric-Guided Agent for Explicit Evidence Sufficiency Verificationin Long-Video Question Answeringcited by 1
  18. 2026Statistical Adversaries: Natural Backdoor-like Adversarial Features in Clean Vision Datasetscited by 0
  19. 2026ReflectWorld-MM: An Entity-Oriented Multimodal Memory System for Open-Ended Video Streamscited by 0
  20. 2026ABot-AgentOS: A General Robotic Agent OS with Lifelong Multi-modal Memorycited by 0
  21. 2026Search over the Visual World: Persistent Visual Memory, Layered Indexes, and Source-Grounded Evidencecited by 0
  22. 2026EgoCITE: Context-Augmented Indexing and Time-Aware Retrieval for Long-Horizon Egocentric Memorycited by 0
  23. 2026Budget-Constrained Embodied Perception: Four Resource Walls and a Pre-Registered Evaluation of Access-Structured Perception on Open Models at less than 31Bcited by 0
  24. 2026From Intent to Evidence: Policy-Steered Multi-Strategy Retrieval for Long-Video Agentscited by 0
  25. 2026EM^2Mem: Event-Centric Multimodal Memory for Large Language Modelscited by 0
  26. 2026ICM-Bench: Person-Level Identity Reasoning in Multimodal Agents with Long-Term Memorycited by 0
  27. 2026CAM: Question Answering on Entity-Centric Videos with Continuous Extraction and Adaptive Queryingcited by 0
  28. 2026Companion-style QA Assistance in Ego-Visioncited by 0
  29. 2026Collaborative Memory for Multi-Agent VLM Systemscited by 0
  30. 2026Agentic Video Understanding: A Surveycited by 0
  31. 2026Retrieved but Not Delivered: Multimodal Memory Delivery for Long-Term Agentscited by 0
  32. 2026Simple Agentic Memory for Generalist Robot Policiescited by 0
  33. 2026LazySloth: Bounded LLM-based Lazy Tree Search for Fast Long Video Comprehensioncited by 0
  34. 2026VideoLoop: Looped Working Memory Against Semantic Thrashing in Long-Form Video Agentscited by 0
  35. 2026Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographiescited by 0
  36. 2026LongEmo: Towards Emotion Understanding and Reasoning in Long Videoscited by 0
  37. 2026MemLife: Curating and Reasoning over Long-Term Egocentric Video Memoriescited by 0
  38. 2026NextMe-800: Anticipating Personal Behavior from Months of Egocentric Videocited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026PyraVid: Hierarchical Multimodal Memory for Long-Horizon Video Reasoning
  2. 2026VideoTapestry: Query-Adaptive Memory Refinement for Multi-Agent Long-Video Understanding
  3. 2024WorldQA: Multimodal World Knowledge in Videos through Long-Chain Reasoning
  4. 2026Homer: Understanding Long-form Videos with Hierarchical Memory and Agentic Reasoning
  5. 2026Reasoning with Memory: A Temporal Granularity-Adaptive Framework for Training-Free Long Video Understanding
  6. 2026MemLens: Benchmarking Multimodal Long-Term Memory in Large Vision-Language Models
  7. 2026Temporal Tree of Thought: Reasoning-Guided Visual Cue Search for Long-Video Understanding
  8. 2026VimRAG: Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph