Research map / Vision-language and multimodal / Visual question answering
MedMKEB: A Comprehensive Knowledge Editing Benchmark for Medical Multimodal Large Language Models
Dexuan Xu, Jieyi Wang, Zhongyan Chai, Yongzhi Cao, Hanpin Wang, Huamin Zhang, Yu Huang
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 |
|---|---|---|---|
| use MEND instead of fine-tuning/other methods | LLaVA · MedMKEB single editing | I-Locality: 10.19-23.02 -> 91.77 | Improved |
| apply SERAC to medical MLLM instead of general MLLM | Biomed-Qwen2-VL · MedMKEB single editing | Reliability: 100.0 -> 35.97 (LLaVA vs Biomed-Qwen2-VL) | Hurt |
| apply IKE to LLaVA-Med | LLaVA-Med · MedMKEB single editing | Reliability: 100.0 -> 7.86 (LLaVA vs LLaVA-Med) | Hurt |
| apply KE (hypernetwork) for portability | general MLLMs · MedMKEB single editing | Portability: 62.35 (LLaVA best) | Improved |
| evaluate portability instead of generality | all editing methods · MedMKEB single editing | Portability: e.g. LLaVA SERAC 99.90 T-Generality -> 57.09 Portability | Hurt |
| apply prompt injection attacks (robustness test) | existing editing methods · MedMKEB | Robustness: e.g. Biomed-Qwen2-VL KE 59.68 -> 4.17 | Hurt |
| fine-tune last LLM layer (FT-LLM) under adversarial prompts | FT-LLM · MedMKEB | Robustness: e.g. Biomed-Qwen2-VL 100.0 -> 99.29 | No clear effect |
| increase number of hops from 1 to 3 | editing methods · MedMKEB multi-hop portability | relative portability change | Hurt |
| use IKE for multi-hop portability | IKE · MedMKEB multi-hop portability | relative portability change | Improved |
| increase sequential editing gap (10 to 100) | FT · MedMKEB sequential editing | reliability and locality | Hurt |
| use SERAC for sequential editing | SERAC · MedMKEB sequential editing | all metrics | Improved |
Closest papers by meaning
- MultiMedBench: A Scenario-Aware Benchmark for Evaluating Knowledge Editing in Medical VQA (AAAI 2026 · similarity 0.92)
- Towards a Multimodal Large Language Model with Pixel-Level Insight for Biomedicine (AAAI 2025 · similarity 0.90)
- MMKE-Bench: A Multimodal Editing Benchmark for Diverse Visual Knowledge (ICLR 2025 · similarity 0.89)
- MedMO: Grounding and Understanding Multimodal Large Language Model for Medical Images (CVPR 2026 · similarity 0.89)
- GMAI-VL & GMAI-VL-5.5M: A Large Vision-Language Model and a Comprehensive Multimodal Dataset Towards General Medical AI (AAAI 2026 · similarity 0.89)
- OmniMedVQA: A New Large-Scale Comprehensive Evaluation Benchmark for Medical LVLM (CVPR 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.
