Research map / Label-efficient and robust learning / Recognition and long-tail
Rep-MTL: Unleashing the Power of Representation-level Task Saliency for Multi-Task Learning
Zedong Wang, Siyuan Li, Dan Xu
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 |
|---|---|---|---|
| add cross-task saliency alignment (CSA) to EW (ablation) | Rep-MTL · NYUv2 | task-level gain: -1.78 -> +1.06 | Improved |
| add task-specific saliency regulation (TSR) to EW (ablation) | Rep-MTL · NYUv2 | task-level gain: -1.78 -> +0.23 | Improved |
| add TSR on top of CSA (ablation) | Rep-MTL · NYUv2 | task-level gain: +1.06 -> +1.70 | Improved |
| add CSA on top of TSR (ablation) | Rep-MTL · NYUv2 | task-level gain: +0.23 -> +1.70 | Improved |
| add CSA to EW (ablation) | Rep-MTL · Cityscapes | task-level gain: -2.05 -> +0.21 | Improved |
| add TSR to EW alone (ablation) | Rep-MTL · Cityscapes | task-level gain: -2.05 -> -0.34 | Improved |
| add both CSA and TSR to EW (ablation) | Rep-MTL · Cityscapes | task-level gain: -2.05 -> +0.62 | Improved |
| use Rep-MTL regularization instead of plain EW | EW · NYUv2 | task-level gain: -1.78 -> +1.70 | Improved |
| use Rep-MTL instead of DB-MTL | DB-MTL · NYUv2 | task-level gain: +1.15 -> +1.70 | Improved |
| use Rep-MTL instead of DB-MTL | DB-MTL · Cityscapes | task-level gain: +0.20 -> +0.62 | Improved |
| use Rep-MTL instead of EW | EW · Office-Home | average accuracy: 78.17 -> 79.14 | Improved |
| use Rep-MTL instead of EW | EW · Office-31 | task-level gain: -0.61 -> +1.31 | Improved |
Closest papers by meaning
- Towards Consistent Multi-Task Learning: Unlocking the Potential of Task-Specific Parameters (CVPR 2025 · similarity 0.90)
- Selective Task Group Updates for Multi-Task Optimization (ICLR 2025 · similarity 0.88)
- Quantifying Task Priority for Multi-Task Optimization (CVPR 2024 · similarity 0.87)
- Representation Surgery for Multi-Task Model Merging (ICML 2024 · similarity 0.87)
- Delve into the Applicability of Advanced Optimizers for Multi-Task Learning (NeurIPS 2026 · similarity 0.87)
- NTKMTL: Mitigating Task Imbalance in Multi-Task Learning from Neural Tangent Kernel Perspective (NeurIPS 2025 · similarity 0.86)
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.
