Research map / Label-efficient and robust learning / Recognition and long-tail
Training Complex Models with Multi-Task Weak Supervision
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 MeTaL to combine multi-task weak supervision sources | End multi-task model · NER | accuracy: 76.9 ± 2.6 -> 82.2 ± 0.8 | Helped |
| Use MeTaL to combine multi-task weak supervision sources | End multi-task model · RE | accuracy: 43.9 ± 2.6 -> 56.7 ± 2.1 | Helped |
| Use MeTaL to combine multi-task weak supervision sources | End multi-task model · Doc | accuracy: 74.2 ± 1.2 -> 76.6 ± 0.4 | Helped |
| Use MeTaL instead of data programming | End multi-task model · NER | accuracy: 78.4 ± 1.2 -> 82.2 ± 0.8 | Helped |
| Use MeTaL instead of data programming | End multi-task model · RE | accuracy: 49.0 ± 2.7 -> 56.7 ± 2.1 | Helped |
| Use MeTaL instead of data programming | End multi-task model · Doc | accuracy: 75.8 ± 0.9 -> 76.6 ± 0.4 | Helped |
| Add class-conditional modeling of unipolar sources (ablation) | MeTaL · Fine-grained classification (three datasets) | accuracy | Helped |
| Jointly model tasks rather than estimating source accuracies for each task separately (ablation) | MeTaL · Fine-grained classification (three datasets) | accuracy: +1.3 points | Helped |
| Train an end model on label-model predictions (ablation) | MeTaL · Fine-grained classification (three datasets) | accuracy | Helped |
| Train an end model on label-model predictions (ablation) | MeTaL · Dataset with the most unlabeled data (n=63k) | accuracy: 7 points | Helped |
Closest papers by meaning
- Learning Dependency Structures for Weak Supervision Models (ICML 2019 · similarity 0.91)
- End-to-End Weak Supervision (NeurIPS 2021 · similarity 0.90)
- Learning from weak labelers as constraints (ICLR 2025 · similarity 0.90)
- A General Framework for Learning from Weak Supervision (ICML 2024 · similarity 0.89)
- Weak Supervision Performance Evaluation via Partial Identification (NeurIPS 2024 · similarity 0.88)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.88)
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.