Research map / Vision-language and multimodal / Visual question answering
DigimonGPT: An Evolvable Agent with Hierarchical Human-like Memory for Video Question Answering
Borui Li, Xingcai Zhang, Tianen Liu, Shuai Wang, Yun Cheng, Shuai Wang
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 semantic memory to episodic memory (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 55.06 -> 60.67 | Improved |
| add semantic memory to episodic memory (ablation) | DigimonGPT · NExT-QA (Descriptive) | accuracy: 50.00 -> 43.75 | Hurt |
| add procedural memory to episodic memory (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 55.06 -> 58.43 | Improved |
| add semantic memory on top of episodic and procedural memory (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 58.43 -> 67.42 | Improved |
| add procedural memory on top of episodic and semantic memory (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 60.67 -> 67.42 | Improved |
| add memory replay versus no replay (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 55.06 -> 67.42 | Improved |
| use dynamic complexity-aware replay instead of fixed Ms=3, Mp=2 (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 65.17 -> 67.42 | Improved |
| replay 5 memories of each type instead of dynamic amount (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 67.42 -> 57.30 | Hurt |
| replay only 1 memory of each type instead of dynamic amount (ablation) | DigimonGPT · NExT-QA (Descriptive) | accuracy: 68.75 -> 43.75 | Mixed |
| increase beta from 0.2 to 0.3 (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 67.42 -> 62.92 | Hurt |
| increase alpha from 0.03 to 0.05 (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 67.42 -> 66.29 | Hurt |
| use normalized hyperparameters alpha=0.11, beta=0.89 (ablation) | DigimonGPT · NExT-QA (All) | accuracy: 67.42 -> 58.43 | Hurt |
Closest papers by meaning
- Seeing, Listening, Remembering, and Reasoning: A Multimodal Agent with Long-Term Memory (ICLR 2026 · similarity 0.89)
- Enhancing Video-LLM Reasoning via Agent-of-Thoughts Distillation (CVPR 2025 · similarity 0.87)
- BDIQA: A New Dataset for Video Question Answering to Explore Cognitive Reasoning through Theory of Mind (AAAI 2024 · similarity 0.87)
- Language-aware Visual Semantic Distillation for Video Question Answering (CVPR 2024 · similarity 0.86)
- Searching Videos as Trees: Self-Correcting Agents for Grounded Long Video QA (NeurIPS 2026 · similarity 0.86)
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.86)
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.
