Research map / Vision-language and multimodal / Visual question answering
Multi-Modal Answer Validation for Knowledge-Based VQA
Jialin Wu, Jiasen Lu, Ashish Sabharwal, Roozbeh Mottaghi
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 answer-guided retrieval from Wikipedia and ConceptNet | MAVEx on ViLBERT · OK-VQA v1.1 test | VQA soft score: 39.45 | Improved |
| Add Google Images as a knowledge source to Wikipedia and ConceptNet (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 39.45 -> 40.28 | Improved |
| Add answer validation to the retrieved-knowledge model (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 39.2 to 40.28 | Improved |
| Add Wikipedia sentence decontextualization (ablation) | MAVEx with Wikipedia knowledge · OK-VQA v1.1 test | VQA soft score: 38.21 -> 38.63 | Improved |
| Add ConceptNet to Wikipedia knowledge (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 38.63 -> 39.45 | Improved |
| Add images to ConceptNet knowledge (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 38.56 -> 39.60 | Improved |
| Add images to Wikipedia knowledge (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 38.63 -> 39.37 | Improved |
| Add Wikipedia to ConceptNet and images (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 39.60 -> 40.28 | Improved |
| Add ConceptNet to Wikipedia and images (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 39.37 -> 40.28 | Improved |
| Replace noun-phrase-level multi-head attention with average pooling (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 39.77 | Hurt |
| Replace question-level multi-head attention with average pooling (ablation) | MAVEx · OK-VQA v1.1 test | VQA soft score: 39.60 | Hurt |
| Expand the answer candidate set from top 5 to top 6 | ViLBERT candidate generation · OK-VQA test | VQA soft score of best answer in candidate set: 59.7 -> 62.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MUTAN — MUTAN: Multimodal Tucker Fusion for Visual Question Answering (ICCV 2017) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FVQA: Fact-Based Visual Question Answering (TPAMI 2017 · similarity 0.90)
- Prompting Large Language Models With Answer Heuristics for Knowledge-Based Visual Question Answering (CVPR 2023 · similarity 0.90)
- Identifying and Resolving Pitfalls of Knowledge-Based VQA Benchmarks: Auditing, Repairing, and Augmenting (ECCV 2026 · similarity 0.90)
- Prophet: Prompting Large Language Models With Complementary Answer Heuristics for Knowledge-Based Visual Question Answering (TPAMI 2025 · similarity 0.90)
- KVQA: Knowledge-Aware Visual Question Answering (AAAI 2019 · similarity 0.90)
- Augmenting Multimodal LLMs with Self-Reflective Tokens for Knowledge-based Visual Question Answering (CVPR 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.
