Research map / Video / Video-language understanding
WorldMM: Dynamic Multimodal Memory Agent for Long Video Reasoning
Woongyeong Yeo, Kangsan Kim, Jaehong Yoon, Sung Ju Hwang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add visual memory to episodic+semantic memory (ablation) | WorldMM-GPT · Average of 5 benchmarks | accuracy (%): 66.8 -> 69.5 | Improved |
| add semantic memory to episodic+visual memory (ablation) | WorldMM-GPT · Average of 5 benchmarks | accuracy (%): 66.9 -> 69.5 | Improved |
| add semantic memory (HabitInsight, Ego-R1 Bench) (ablation) | WorldMM-GPT · Ego-R1 Bench HabitInsight | accuracy (%): 53.9 -> 76.9 | Improved |
| add episodic memory to visual-only memory (ablation) | WorldMM-GPT · Average of 5 benchmarks | accuracy (%): 44.9 -> 64.9 | Improved |
| use multi-scale episodic memory instead of a single fixed timescale (ablation) | WorldMM-8B · EgoLifeQA | accuracy (%): 51.8 -> 56.4 | Improved |
| use graph retrieval instead of embedding retrieval for episodic memory (ablation) | WorldMM-8B · EgoLifeQA | accuracy (%): 52.0 -> 56.4 | Improved |
| add semantic memory consolidation (ablation) | WorldMM-8B · EgoLifeQA | accuracy (%): 53.0 -> 56.4 | Improved |
| use both feature-based and timestamp-based visual retrieval (ablation) | WorldMM-8B · LVBench | accuracy (%): 52.4 -> 55.4 | Improved |
| increase max retrieval iterations from 1 to 5 (ablation) | WorldMM-GPT · EgoLifeQA | accuracy (%) | Improved |
| WorldMM-GPT vs GPT-5 base model | GPT-5 · Average of 5 benchmarks | accuracy (%): 61.1 -> 69.5 | Improved |
| WorldMM-GPT vs GPT-5 on LVBench | GPT-5 · LVBench | accuracy (%): 60.4 -> 61.9 | Improved |
| WorldMM-8B vs Qwen3-VL-8B on HippoVlog | Qwen3-VL-8B · HippoVlog | accuracy (%): 74.4 -> 69.7 | Hurt |
Closest papers by meaning
- MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding (CVPR 2024 · similarity 0.90)
- Navigating Massive Visual Context in Retrieval-Augmented Generation via Multimodal Memory Graph (ICML 2026 · similarity 0.90)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.90)
- Temporal Reasoning Transfer from Text to Video (ICLR 2025 · similarity 0.90)
- MMR-V: What's Left Unsaid? A Benchmark for Multimodal Deep Reasoning in Videos (ICLR 2026 · similarity 0.89)
- MMOU: A Massive Multi-Task Omni Understanding and Reasoning Benchmark for Long and Complex Real-World Videos (NeurIPS 2026 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
