Research map / Architectures and efficiency / Pruning and efficient tuning
Attentive Single-Tasking of Multiple Tasks
Check 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 |
|---|---|---|---|
| Add task-specific SE modulation (ablation) | Five-task R-26 multi-task network · PASCAL Edge | Edge: 69.2 -> 71.1 | Helped |
| Add task-specific SE modulation (ablation) | Five-task R-26 multi-task network · PASCAL Seg | Seg: 60.20 -> 64.00 | Helped |
| Use task-specific residual adapters instead of shared task processing (ablation) | Five-task R-26 multi-task network · PASCAL Edge | Edge: 69.2 -> 70.5 | Helped |
| Add adversarial task-gradient training (ablation) | Five-task R-26 multi-task network without modulation · PASCAL Seg | Seg: 60.20 -> 62.20 | Helped |
| Add adversarial task-gradient training (ablation) | Five-task R-26 network with SE modulation · PASCAL Parts | Parts: 56.84 -> 57.25 | Helped |
| Modulate the decoder only rather than both encoder and decoder (ablation) | Five-task R-26 SE-modulated network · PASCAL Edge | Edge: 70.6 -> 71.1 | No clear effect |
| Add SE modulation and adversarial training (ablation) | Four-task R-50 multi-task network · NYUD-v2 Depth | Depth: 0.70 -> 0.57 | Helped |
| Add SE modulation and adversarial training (ablation) | Three-task R-50 multi-task network · FSV Disp | Disp: 0.078 -> 0.063 | Helped |
| Use SE blocks in the backbone (ablation) | Single-task R-26 network · PASCAL Edge | Edge: 70.3 -> 71.3 | Helped |
| Use SE blocks in the backbone (ablation) | Five-task R-26 network · PASCAL Normals | Norm: 16.68 -> 17.04 | Hurt |
Closest papers by meaning
- Provable Multi-Task Representation Learning by Two-Layer ReLU Neural Networks (ICML 2024 · similarity 0.85)
- Quantifying Task Priority for Multi-Task Optimization (CVPR 2024 · similarity 0.85)
- Selective Task Group Updates for Multi-Task Optimization (ICLR 2025 · similarity 0.84)
- Inductive biases of multi-task learning and finetuning: multiple regimes of feature reuse (NeurIPS 2024 · similarity 0.83)
- Task Singular Vectors: Reducing Task Interference in Model Merging (CVPR 2025 · similarity 0.83)
- DMTG: One-Shot Differentiable Multi-Task Grouping (ICML 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.