Research map / Video / Video-language understanding
ViG-RAG: Video-aware Graph Retrieval-Augmented Generation via Temporal and Semantic Hybrid Reasoning
Zongsheng Cao, Anran Liu, Yangfan He, Jing Li, Bo Zhang, Zigan 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 ViG-RAG retrieval to Video-LLaVA | Video-LLaVA 7B · Video-MME | overall accuracy: 39.6 -> 43.5 | Improved |
| add ViG-RAG retrieval to LLaVA-NeXT-Video | LLaVA-NeXT-Video 7B · Video-MME | overall accuracy: 43.0 -> 54.2 | Improved |
| add ViG-RAG retrieval to LongVA | LongVA 7B · Video-MME | overall accuracy: 51.4 -> 58.8 | Improved |
| add ViG-RAG retrieval to Long-LLaVA | Long-LLaVA 7B · Video-MME | overall accuracy: 52.0 -> 57.7 | Improved |
| add ViG-RAG retrieval to Qwen2-VL | Qwen2-VL 72B · Video-MME | overall accuracy: 64.9 -> 72.4 | Improved |
| add ViG-RAG retrieval to LLaVA-Video | LLaVA-Video 72B · Video-MME | overall accuracy: 67.1 -> 74.4 | Improved |
| add ViG-RAG retrieval to Qwen2-VL, long videos | Qwen2-VL 72B · Video-MME (long) | accuracy: 56.3 -> 72.2 | Improved |
| ViG-RAG instead of VideoRAG in win-rate comparison vs GraphRAG-l | ViG-RAG · LongerVideos (overall) | Overall win rate: 35.81% -> 41.77% | Improved |
| add probabilistic temporal knowledge graph (PTKG) (ablation) | ViG-RAG · benchmarks (Figure 3) | retrieval accuracy / overall performance | Improved |
| add query enhancement (ablation) | ViG-RAG · benchmarks (Figure 3) | retrieval accuracy / overall performance | Improved |
| add GMM-based adaptive filtering (ablation) | ViG-RAG · benchmarks (Figure 3) | retrieval accuracy / overall performance | Improved |
Closest papers by meaning
- Vgent: Graph-based Retrieval-Reasoning-Augmented Generation For Long Video Understanding (NeurIPS 2025 · similarity 0.94)
- AdaVideoRAG: Omni-Contextual Adaptive Retrieval-Augmented Efficient Long Video Understanding (NeurIPS 2025 · similarity 0.93)
- Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension (NeurIPS 2025 · similarity 0.91)
- Re-ranking Reasoning Context with Tree Search Makes Large Vision-Language Models Stronger (ICML 2025 · similarity 0.88)
- Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames (NeurIPS 2025 · similarity 0.88)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.88)
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.
