Research map / Architectures and efficiency / Pruning and efficient tuning
Learning Sparse Sharing Architectures for Multiple 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 sparse sharing instead of hard sharing | Multi-task sequence labeling · CoNLL-2003 NER | Test F1: 88.68 -> 90.35 | Helped |
| Use sparse sharing instead of hard sharing | Multi-task sequence labeling · OntoNotes 5.0 NER | Test F1: 82.95 -> 83.42 | Helped |
| Use sparse sharing instead of Meta-MTL-LSTM | Exp3 multi-task sequence labeling · PTB POS | Test score: 97.45 -> 97.57 | Helped |
| Use sparse sharing instead of hard sharing | NER and synthetic position prediction · CoNLL-2003 NER | Test F1: 61.62 -> 71.46 | Helped |
| Add multi-task warmup (ablation) | Sparse sharing · CoNLL-2003 NER | Test F1: 89.62 -> 90.35 | Helped |
| Add multi-task warmup (ablation) | Sparse sharing · CoNLL-2003 POS | Test accuracy: 95.36 -> 95.56 | Helped |
| Add multi-task warmup (ablation) | Sparse sharing · CoNLL-2003 Chunking | Test F1: 91.04 -> 91.55 | Helped |
| Add multi-task warmup (ablation) | Sparse sharing · OntoNotes 5.0 NER | Test F1: 81.15 -> 83.42 | Helped |
| Add multi-task warmup (ablation) | Sparse sharing · OntoNotes 5.0 POS | Test accuracy: 97.53 -> 97.54 | No clear effect |
| Add multi-task warmup (ablation) | Sparse sharing · OntoNotes 5.0 Chunking | Test F1: 95.48 -> 95.56 | No clear effect |
| Use task-specific pruned subnet instead of a dense single-task model (ablation) | Single-task sequence labeling · CoNLL-2003 POS | Test accuracy: 95.09 -> 95.11 | No clear effect |
| Use task-specific pruned subnet instead of a dense single-task model (ablation) | Single-task sequence labeling · CoNLL-2003 NER | Test F1: 89.36 -> 89.39 | No clear effect |
Closest papers by meaning
- Deep Multi-task Representation Learning: A Tensor Factorisation Approach (ICLR 2017 · similarity 0.88)
- Learning Multi-Task Sparse Representation Based on Fisher Information (AAAI 2024 · similarity 0.88)
- Cross-Stitch Networks for Multi-Task Learning (CVPR 2016 · similarity 0.86)
- Sharing Knowledge in Multi-Task Deep Reinforcement Learning (ICLR 2020 · similarity 0.86)
- Task-aware Orthogonal Sparse Network for Exploring Shared Knowledge in Continual Learning (ICML 2024 · similarity 0.85)
- Random Sparse Lifts: Construction, Analysis and Convergence of finite sparse networks (ICLR 2024 · 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.