Research map / Vision-language and multimodal / Cross-modal learning
LLaVA-MS-PIT: Multi-Modal Schema-Guided Progressive Instruction Tuning for Multi-Modal Event Extraction
Hui Zhang, Po Hu, Wei Emma Zhang
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 |
|---|---|---|---|
| data cleaning of imSitu (ODT to CDT) (ablation) | LLaVA-MS-PIT · M2E2 visual events: event mention | F1: 60.7 -> 72.3 | Improved |
| data cleaning of imSitu (ODT to CDT) (ablation) | LLaVA-MS-PIT · M2E2 visual events: argument role | F1: 34.2 -> 51.9 | Improved |
| add textual event schema tuning (ablation) | CDT · M2E2 multimedia events: event mention | F1: 81.8 -> 85.1 | Improved |
| add visual event schema tuning (ablation) | CDT · M2E2 multimedia events: event mention | F1: 81.8 -> 86.5 | Improved |
| combine textual and visual schema tuning (ablation) | LLaVA-MS-PIT · M2E2 multimedia events: argument role | F1: 42.5 -> 46.8 | Improved |
| add textual schema tuning (ablation) | CDT · ACE 2005 argument role | F1: 55.5 -> 58.0 | Improved |
| add textual schema tuning (ablation) | CDT · M2E2 text argument role | F1: 36.0 -> 36.0 | No clear effect |
| add visual schema tuning on visual argument role precision/recall (ablation) | CDT · M2E2 text argument role | F1: 35.4 vs 36.0 | Mixed |
| LLaVA-MS-PIT vs X-MTL on text argument role | LLaVA-1.5-7B · M2E2 text events: argument role | F1: 36.0 -> 36.2 (loose), 31.1 (strict) | Hurt |
| LLaVA-MS-PIT vs X-MTL on multimedia event mention | LLaVA-1.5-7B · M2E2 multimedia events: event mention | F1: 66.2 -> 87.5 | Improved |
Closest papers by meaning
- Cross-modal Multi-task Learning for Multimedia Event Extraction (AAAI 2025 · similarity 0.86)
- Adapt-$\infty$: Scalable Continual Multimodal Instruction Tuning via Dynamic Data Selection (ICLR 2025 · similarity 0.85)
- SCHEMA: State CHangEs MAtter for Procedure Planning in Instructional Videos (ICLR 2024 · similarity 0.85)
- LongVALE: Vision-Audio-Language-Event Benchmark Towards Time-Aware Omni-Modal Perception of Long Videos (CVPR 2025 · similarity 0.84)
- Can Multi-Modal LLMs Provide Live Step-by-Step Task Guidance? (NeurIPS 2025 · similarity 0.84)
- DanmakuTPPBench: A Multi-modal Benchmark for Temporal Point Process Modeling and Understanding (NeurIPS 2025 · similarity 0.84)
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.
