Research map / Vision-language and multimodal / Visual question answering
Attention-Based Context Aware Reasoning for Situation Recognition
Thilini Cooray, Ngai-Man Cheung, Wei Lu
Check 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 attention-weighted context from neighboring roles to each role query (CAQ) (ablation) | TDA frame recognition · imSitu development | value accuracy: 72.96 -> 73.62 | Improved |
| Add attention-weighted context from neighboring roles to each role query (CAQ) (ablation) | TDA frame recognition · imSitu development | value-all accuracy: 37.60 -> 38.71 | Improved |
| Use context to gate image regions (CAI) rather than incorporating it into the query (ablation) | TDA frame recognition · imSitu development | value accuracy: 72.96 -> 73.17 | Improved |
| Use context to gate image regions (CAI) rather than incorporating it into the query (ablation) | TDA frame recognition · imSitu development | value-all accuracy: 37.60 -> 37.95 | Improved |
| Add an auxiliary image reconstruction loss (CAIR) (ablation) | TDA frame recognition · imSitu development | value accuracy: 72.96 -> 73.30 | Improved |
| Add an auxiliary image reconstruction loss (CAIR) (ablation) | TDA frame recognition · imSitu development | value-all accuracy: 37.60 -> 38.17 | Improved |
| Combine query-context CAQ and image-context CAI (ablation) | CAQ frame recognition · imSitu development | value accuracy: 73.62 -> 72.94 | Hurt |
| Combine query-context CAQ and image-context CAI (ablation) | CAQ frame recognition · imSitu development | value-all accuracy: 38.71 -> 37.38 | Hurt |
| Add CAIR to CAQ (ablation) | CAQ frame recognition · imSitu development | value accuracy: 73.62 -> 73.62 | No clear effect |
| Add CAIR to CAQ (ablation) | CAQ frame recognition · imSitu development | value-all accuracy: 38.71 -> 38.63 | Hurt |
| Use RE-VGG verb scores alongside CAQ | CAQ situation recognition · imSitu test | top-1 verb accuracy: 36.97 -> 38.19 | Improved |
| Use RE-VGG verb scores alongside CAQ | CAQ situation recognition · imSitu test | top-1 value accuracy: 29.29 -> 30.23 | Improved |
Closest papers by meaning
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.85)
- Semantic-Enriched Latent Visual Reasoning (ICML 2026 · similarity 0.84)
- Identifying and Mitigating Position Bias of Multi-image Vision-Language Models (CVPR 2025 · similarity 0.84)
- Not Another Text Benchmark: Putting the “Visual" Back in Visual Question Answering for Large Video Models (NeurIPS 2026 · similarity 0.84)
- VisRes Bench: On Evaluating the Visual Reasoning Capabilities of VLMs (CVPR 2026 · similarity 0.84)
- ConTextual: Evaluating Context-Sensitive Text-Rich Visual Reasoning in Large Multimodal Models (ICML 2024 · similarity 0.83)
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.
