Research map / Architectures and efficiency / Attention and transformers
MTFormer: Multi-task Learning via Transformer and Cross-Task Reasoning
Xiaogang Xu, Hengshuang Zhao, Vibhav Vineet, Ser-Nam Lim, Antonio Torralba
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 attention (ablation) | MTFormer (no CL) · NYUD-v2 | delta_m %: 2.87 -> 3.56 | Improved |
| add cross-task attention (ablation) | MTFormer (no CL) · PASCAL-Context | delta_m %: 1.55 -> 1.91 | Improved |
| add cross-task contrastive learning (ablation) | MTFormer · NYUD-v2 | delta_m %: 3.56 -> 4.12 | Improved |
| add cross-task contrastive learning (ablation) | MTFormer · PASCAL-Context | delta_m %: 1.91 -> 2.41 | Improved |
| add feature mapping before MLP fusion (ablation) | MTFormer (no CL) · NYUD-v2 | delta_m %: 3.49 -> 3.56 | Improved |
| add feature mapping before MLP fusion (ablation) | MTFormer (no CL) · PASCAL-Context | delta_m %: 1.58 -> 1.91 | Improved |
| add MLP fusion instead of summation (ablation) | MTFormer (no CL) · NYUD-v2 | delta_m %: 3.38 -> 3.56 | Improved |
| add feature balance (smaller cross-task channels) (ablation) | MTFormer (no CL) · PASCAL-Context | delta_m %: 1.70 -> 1.91 | Improved |
| use transformer instead of CNN for multi-task learning | MTL · NYUD-v2 | delta_m %: -2.61 -> +2.87 | Improved |
| use transformer instead of CNN for multi-task learning | MTL · PASCAL-Context | delta_m %: -4.96 -> +1.55 | Improved |
| share a deep decoder with lightweight branches (MTL vs STL params) | Transformer MTL · NYUD-v2 | Params (M): 114.26 -> 62.45 | Improved |
| add MTL with few-shot part annotations | MTFormer · PASCAL few-shot (1% part labels) | Human part mIoU: 11.27 -> 51.74 | Improved |
Closest papers by meaning
- Resolving Token-Space Gradient Conflicts: Token Space Manipulation for Transformer-Based Multi-Task Learning (ICCV 2025 · similarity 0.89)
- TADFormer: Task-Adaptive Dynamic TransFormer for Efficient Multi-Task Learning (CVPR 2025 · similarity 0.89)
- Transforming Vision Transformer: Towards Efficient Multi-Task Asynchronous Learner (NeurIPS 2024 · similarity 0.88)
- MTLoRA: Low-Rank Adaptation Approach for Efficient Multi-Task Learning (CVPR 2024 · similarity 0.87)
- Mr. DETR: Instructive Multi-Route Training for Detection Transformers (CVPR 2025 · similarity 0.86)
- Task-Adaptive Prompted Transformer for Cross-Domain Few-Shot Learning (AAAI 2024 · 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.
