Research map / Label-efficient and robust learning / Recognition and long-tail
Task Prototype-Based Knowledge Retrieval for Multi-Task Learning from Partially Annotated Data
Youngmin Oh, Hyung-Il Kim, Jung Uk Kim
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 vector quantization with TAE loss (ablation) | prototype-based knowledge retrieval MTL · PASCAL-Context one-label | Semseg mIoU: 44.73 -> 44.83 | No clear effect |
| add task knowledge embedding (TKE) loss (ablation) | prototype-based knowledge retrieval MTL · PASCAL-Context one-label | Semseg mIoU: 44.83 -> 58.21 | Improved |
| add task consistency (TC) loss (ablation) | prototype-based knowledge retrieval MTL · PASCAL-Context one-label | Semseg mIoU: 58.21 -> 59.78 | Improved |
| add TKE loss (ablation) | prototype-based knowledge retrieval MTL · PASCAL-Context one-label | Normal mErr: 16.22 -> 15.67 | Improved |
| increase task prototype slot dimension from 1024 to 2048 (ablation) | prototype-based knowledge retrieval MTL · NYUD-v2 one-label | Semseg mIoU: 45.95 -> 45.33 | Hurt |
| reduce slot dimension from 1024 to 256 (ablation) | prototype-based knowledge retrieval MTL · NYUD-v2 one-label | Semseg mIoU: 45.95 -> 44.91 | Hurt |
| use proposed method instead of DiffusionMTL (Feature) | prototype-based knowledge retrieval MTL · PASCAL-Context one-label | Semseg mIoU: 57.78 -> 59.78 | Improved |
| use proposed method instead of DiffusionMTL (Feature) | prototype-based knowledge retrieval MTL · PASCAL-Context one-label | Boundary odsF: 64.50 -> 65.10 | Improved |
| use proposed method instead of TaskPrompter | prototype-based knowledge retrieval MTL · NYUD-v2 one-label ViT-L | Semseg mIoU: 48.68 -> 50.08 | Improved |
| use proposed method instead of DiffusionMTL (Prediction) | prototype-based knowledge retrieval MTL · PASCAL-Context ResNet-50 | Semseg mIoU: 60.92 -> 62.23 | Improved |
| proposed method vs single-task learning | prototype-based knowledge retrieval MTL · NYUD-v2 random-label | Normal mErr: 24.65 vs 24.67 | Mixed |
Closest papers by meaning
- Amalgamating Multi-Task Models with Heterogeneous Architectures (AAAI 2024 · similarity 0.86)
- Towards Principled Task Grouping for Multi-Task Learning (NeurIPS 2026 · similarity 0.85)
- MTL-UE: Learning to Learn Nothing for Multi-Task Learning (ICML 2025 · similarity 0.85)
- Towards Consistent Multi-Task Learning: Unlocking the Potential of Task-Specific Parameters (CVPR 2025 · similarity 0.84)
- NTKMTL: Mitigating Task Imbalance in Multi-Task Learning from Neural Tangent Kernel Perspective (NeurIPS 2025 · similarity 0.84)
- Provable In-Context Vector Arithmetic via Retrieving Task Concepts (ICML 2025 · similarity 0.84)
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.
