Research map / Probabilistic and scientific modeling / Causal discovery
Generative multitask learning mitigates target-causing confounding
Taro Makino, Krzysztof J. Geras, Kyunghyun Cho
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 |
|---|---|---|---|
| Use GMTL inference with an optimal α instead of conventional DMTL inference | Four multitask learning methods · Attributes of People and Taskonomy, simulated target shifts | accuracy | Improved |
| Increase α as target shift becomes more severe (ablation) | GMTL · Attributes of People and Taskonomy, simulated target shifts | accuracy | Improved |
| Increase α, removing more of the empirical joint-target prior (ablation) | GMTL · In-distribution test sets | accuracy | Hurt |
| Use the target-shift severity metric and α selection on long sleeves as the main task and hat as auxiliary | GMTL · Attributes of People: long sleeves (main), hat (auxiliary), simulated target shifts | accuracy | Mixed |
| Replace shared-trunk multitask learning with no parameter sharing | Taskonomy · Object and scene classification, ID test | top-1 accuracy: 0.710 ± 0.001 -> 0.749 ± 0.001; 0.719 ± 0.001 -> 0.730 ± 0.001 | Improved |
| Use Cross-stitch Networks instead of single-task-equivalent NPS | Attributes of People · Hat (main), long sleeves (auxiliary), ID test | accuracy: 0.917 ± 0.003 -> 0.920 ± 0.001 | Improved |
| Use Cross-stitch Networks instead of single-task-equivalent NPS | Taskonomy · Scene classification (main), object classification (auxiliary), ID test | top-1 accuracy: 0.730 ± 0.001 -> 0.737 ± 0.001 | Improved |
Closest papers by meaning
- Identifying and Mitigating Spurious Correlation in Multi-Task Learning (CVPR 2025 · similarity 0.84)
- Learning to Focus: Causal Attention Distillation via Gradient‐Guided Token Pruning (NeurIPS 2025 · similarity 0.84)
- NTKMTL: Mitigating Task Imbalance in Multi-Task Learning from Neural Tangent Kernel Perspective (NeurIPS 2025 · similarity 0.83)
- Distributional Alignment as a Criterion for Designing Task Vectors in In-Context Learning (NeurIPS 2026 · similarity 0.83)
- Towards Consistent Multi-Task Learning: Unlocking the Potential of Task-Specific Parameters (CVPR 2025 · similarity 0.83)
- Towards Few-Shot Adaptation of Foundation Models via Multitask Finetuning (ICLR 2024 · 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.
