Research map / Label-efficient and robust learning / Recognition and long-tail
A Hierarchical Multi-Task Approach for Learning Embeddings from Semantic Tasks
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 |
|---|---|---|---|
| Use the full hierarchical multi-task model instead of single-task RE | Relation extraction · ACE05 | F1: 55.99 -> 62.69 | Helped |
| Use the full multi-task model instead of single-task NER | NER · OntoNotes 5.0 | F1: 87.12 -> 87.10 | No clear effect |
| Use the full model with gold mentions for coreference evaluation | Full Model · ACE05 CR | Avg. F1: 64.78 -> 70.29* | Helped |
| Add RE supervision to NER and EMD (ablation) | NER + EMD · OntoNotes 5.0 | F1: 86.91 -> 87.51 | Helped |
| Add RE supervision to NER and EMD (ablation) | NER + EMD · ACE05 | F1: 86.02 -> 86.26 | Helped |
| Add CR supervision to NER and EMD (ablation) | NER + EMD · OntoNotes 5.0 | F1: 86.91 -> 87.50 | Helped |
| Add CR supervision to NER and EMD (ablation) | NER + EMD · ACE05 | F1: 86.02 -> 85.87 | Hurt |
| Put EMD below NER in the hierarchy (ablation) | Hierarchical multi-task model · OntoNotes 5.0 | F1: 87.50 -> 86.21 | Hurt |
| Remove ELMo embeddings (ablation) | Full Model - GM · ACE05 | F1: 87.10 -> 84.33; 87.24 -> 83.13; 62.69 -> 57.47; 70.29 -> 66.44 | Hurt |
| Remove character-level embeddings (ablation) | Full Model - GM · ACE05 | F1: 84.33 -> 79.81; 83.13 -> 83.00; 57.47 -> 53.77; 66.44 -> 64.26 | Mixed |
| Use proportional rather than uniform task sampling | Hierarchical multi-task model · Multiple tasks | performance | Helped |
| Train with multiple tasks rather than single-task EMD (ablation) | EMD · ACE05 | F1: 86.14 -> 87.24 | Mixed |
Closest papers by meaning
- Multi-Task Learning of Hierarchical Vision-Language Representation (CVPR 2019 · similarity 0.89)
- Progressive Multi-task Learning with Controlled Information Flow for Joint Entity and Relation Extraction (AAAI 2021 · similarity 0.88)
- An Information-theoretic Multi-task Representation Learning Framework for Natural Language Understanding (AAAI 2025 · similarity 0.88)
- Learning Task-Agnostic Representations through Multi-Teacher Distillation (NeurIPS 2025 · similarity 0.86)
- Provable Multi-Task Representation Learning by Two-Layer ReLU Neural Networks (ICML 2024 · similarity 0.85)
- Banyan: Improved Representation Learning with Explicit Structure (ICML 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.