Research map / Vision-language and multimodal / Visual question answering
Dynamic Multimodal Evaluation via Knowledge-Enhanced Benchmark Evolution
Junzhe Zhang, Huixuan Zhang, Xiaojun Wan
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 |
|---|---|---|---|
| expand questions to 3 hops with KBE-DME | GPT-4o · OK-VQA | accuracy: 50.33 -> 39.19 | Hurt |
| expand questions to 3 hops with KBE-DME | Gemini-2.5-pro · OK-VQA | accuracy: 49.94 -> 32.92 | Hurt |
| expand questions to 3 hops with KBE-DME | Claude · OK-VQA | accuracy: 52.26 -> 37.60 | Hurt |
| expand questions to 3 hops with KBE-DME | LLaVA-OV · OK-VQA | accuracy: 53.46 -> 29.52 | Hurt |
| expand questions to 3 hops with KBE-DME | Qwen-2.5-VL · OK-VQA | accuracy: 47.35 -> 33.04 | Hurt |
| expand questions to 3 hops with KBE-DME | GPT-4o · A-OKVQA | accuracy: 60.13 -> 36.27 | Hurt |
| expand questions to 3 hops with KBE-DME | Gemini-2.5-pro · A-OKVQA | accuracy: 58.99 -> 32.03 | Hurt |
| expand questions to 3 hops with KBE-DME | Claude · A-OKVQA | accuracy: 57.68 -> 32.84 | Hurt |
| expand questions to 3 hops with KBE-DME | LLaVA-OV · A-OKVQA | accuracy: 60.78 -> 25.33 | Hurt |
| expand questions to 3 hops with KBE-DME | Qwen-2.5-VL · A-OKVQA | accuracy: 57.35 -> 30.88 | Hurt |
| increase hops from raw to 3 on question length | OK-VQA questions · OK-VQA | question words: 8.18 -> 18.8 | Improved |
| increase hops from raw to 3 on key subgraph edges | OK-VQA questions · OK-VQA | average |E_K|: 2.98 -> 6.04 | Improved |
Closest papers by meaning
- MDK12-Bench: A Multi-Discipline Benchmark for Evaluating Reasoning in Multimodal Large Language Models (AAAI 2026 · similarity 0.90)
- UPME: An Unsupervised Peer Review Framework for Multimodal Large Language Model Evaluation (CVPR 2025 · similarity 0.90)
- MMBench-Live: A Continuously Evolving Benchmark for Multimodal Models (ICML 2026 · similarity 0.89)
- Dynamic Multimodal Evaluation with Flexible Complexity by Vision-Language Bootstrapping (ICLR 2025 · similarity 0.89)
- MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities (ICML 2024 · similarity 0.89)
- Benchmarking Multimodal Retrieval Augmented Generation with Dynamic VQA Dataset and Self-adaptive Planning Agent (ICLR 2025 · 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.
