Research map / Video / Video-language understanding
Hierarchical Event Grounding
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 hierarchical pretraining | Baseline bi-encoder · Wikipedia multilingual dev/test | Recall@min: 65.4 / 54.8 -> 71.3 / 58.1 | Helped |
| add hierarchical pretraining | Baseline bi-encoder · Wikipedia crosslingual dev/test | Recall@min: 53.8 / 32.8 -> 59.7 / 37.0 | Helped |
| use hierarchical joint learning without pretraining (ablation) | Baseline bi-encoder · Wikipedia multilingual dev/test | Recall@min: 65.4 / 54.8 -> 63.3 / 51.4 | Hurt |
| add joint learning after hierarchical pretraining (ablation) | Baseline bi-encoder · Wikipedia multilingual dev/test | Recall@min: 71.3 / 58.1 -> 67.6 / 55.2 | Mixed |
| use hierarchical joint learning without pretraining (ablation) | Baseline bi-encoder · Wikipedia crosslingual dev/test | Recall@min: 53.8 / 32.8 -> 51.6 / 33.3 | Mixed |
| add joint learning after hierarchical pretraining (ablation) | Baseline bi-encoder · Wikipedia crosslingual dev/test | Recall@min: 59.7 / 37.0 -> 55.8 / 38.8 | Mixed |
| add hierarchical joint learning | Baseline cross-encoder · Wikinews multilingual | Strict accuracy: 51.7 -> 55.8 | Helped |
| add hierarchical joint learning (ablation) | Baseline bi-encoder · Wikinews multilingual | Recall@min: 68.6 -> 67.4 | Hurt |
| add hierarchical pretraining and joint learning | Baseline bi-encoder · Wikinews crosslingual | Recall@min: 51.2 -> 53.7 | Helped |
| use cross-encoder reranking instead of bi-encoder retrieval alone | Best-performing associated systems · Wikipedia multilingual test | strict accuracy | Helped |
| use cross-encoder reranking instead of bi-encoder retrieval alone | Associated systems · Wikipedia non-test cases | strict accuracy | Hurt |
| rank candidate parents by overlap of their linked mention sets | Best-performing bi-encoder retrieval results · Hierarchical relation extraction, multilingual and crosslingual | Recall@4: 69.0 / 72.6; 78.3 / 60.5 | Helped |
Closest papers by meaning
- Beyond Grounding: Extracting Fine-Grained Event Hierarchies across Modalities (AAAI 2024 · similarity 0.89)
- Logic Induced High-Order Reasoning Network for Event-Event Relation Extraction (AAAI 2025 · similarity 0.84)
- CHRONOS: A Schema-Based Event Understanding and Prediction System (AAAI 2024 · similarity 0.83)
- A Hierarchical Network for Multimodal Document-Level Relation Extraction (AAAI 2024 · similarity 0.83)
- EventSum: A Large-Scale Event-Centric Summarization Dataset for Chinese Multi-News Documents (AAAI 2025 · similarity 0.83)
- Just What You Desire: Constrained Timeline Summarization with Self-Reflection for Enhanced Relevance (AAAI 2025 · 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.