Research map / Vision-language and multimodal / Visual question answering
Learning Visual Knowledge Memory Networks for Visual Question Answering
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 knowledge memory to MLB | MLB · VQA v1.0 test-dev, Open-Ended Other | accuracy: 54.9 -> 57.0 | Helped |
| Add triple replication so all triple-part key/value mappings are represented (ablation) | VKMN · VQA v1.0 test-dev, Open-Ended Other | accuracy: 53.9 -> 57.0 | Helped |
| Add visual input to the memory key/value features (ablation) | VKMN · VQA v1.0 test-dev, Open-Ended Other | accuracy: 28.3 -> 57.0 | Helped |
| Add memory blocks for reading knowledge (ablation) | MLB · VQA v1.0 test-dev, Open-Ended Other | accuracy: 54.9 -> 57.0 | Helped |
| Use TransE instead of BoW for knowledge encoding (ablation) | VKMN · VQA v1.0 test-dev, Open-Ended Other | accuracy: 56.1 -> 57.0 | Helped |
| Use VKMN instead of SVA | VQA v1.0 test-standard, Open-Ended Other · VQA v1.0 test-standard, Open-Ended Other | accuracy: 55.9 -> 56.9 | Helped |
| Use VKMN instead of MLB with visual genome data | VQA v2.0 · VQA v2.0 test-standard Other | accuracy: 56.98 -> 57.79 | Helped |
| Ensemble VKMN with SVA | VKMN · VQA v2.0 test-standard | All accuracy: 64.36 -> 66.67 | Helped |
| Use VKMN instead of SVA | VQA v2.0 single-model comparison · VQA v2.0 test-standard | All accuracy: 64.55 -> 64.36 | Hurt |
Closest papers by meaning
- REVIVE: Regional Visual Representation Matters in Knowledge-Based Visual Question Answering (NeurIPS 2022 · similarity 0.92)
- Dynamic Key-Value Memory Enhanced Multi-Step Graph Reasoning for Knowledge-Based Visual Question Answering (AAAI 2022 · similarity 0.92)
- Prompting Large Language Models With Answer Heuristics for Knowledge-Based Visual Question Answering (CVPR 2023 · similarity 0.92)
- OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge (CVPR 2019 · similarity 0.91)
- BOK-VQA: Bilingual outside Knowledge-Based Visual Question Answering via Graph Representation Pretraining (AAAI 2024 · similarity 0.91)
- Separation of Powers: On Segregating Knowledge from Observation in LLM-enabled Knowledge-based Visual Question Answering (CVPR 2025 · similarity 0.91)
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.