Research map / Architectures and efficiency / Pruning and efficient tuning
TDSS: Task Dynamic-Synergistic Skill Adaptation for Boosting Efficient and Scalable Multi-Task Learning in Dense Visual Prediction
Haiming Yao, Qiyu Chen, Wei Luo, Zheng Zhang, Jianxing Liao, Wei You
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 TSAI to the baseline (ablation) | TDSS · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: 0 -> +5.76 | Improved |
| add TDSA to the baseline (ablation) | TDSS · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: 0 -> +0.94 | Improved |
| combine TDSA and TSAI (ablation) | TDSS · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: +5.76 -> +8.37 | Improved |
| add skill experts and dynamic task gating (ablation) | TDSA · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: +7.85 -> +8.37 | Improved |
| insert TDSA in both attention and FFN rather than one (ablation) | TDSS · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: +7.41 / +7.19 -> +8.37 | Improved |
| increase number of skills M from 10 (ablation) | TDSA · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: +8.37 (M=10) vs +8.12 / +8.13 / +8.23 | No clear effect |
| increase low-rank dimension R beyond C/4 (ablation) | TDSA · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: +8.37 -> +8.76 (R=C/2), params 3.52M -> 5.89M | Mixed |
| more synergistic modules K (1 -> 4 -> 6) (ablation) | TSAI · PASCAL-Context (Swin-T, 4 tasks) | Delta_m: +5.29 -> +8.37 -> +9.18 | Improved |
| use lateral topology versus MTLoRA/TADFormer | TDSS · PASCAL-Context (Swin-T, 4 tasks) | GPU memory (GB): 14.65 vs 18.22 / 20.30 | Improved |
| use TDSS rather than MTLoRA | TDSS · PASCAL-Context (Swin-T, 4 tasks) | Training time (s/batch): 0.232 -> 0.380 | Hurt |
| use task-specific scale fusion with global interaction (ablation) | TSAI · PASCAL-Context (Swin-T, 4 tasks) | Delta_m | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VMT-Adapter — VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding (AAAI 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- TADFormer: Task-Adaptive Dynamic TransFormer for Efficient Multi-Task Learning (CVPR 2025 · similarity 0.89)
- VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding (AAAI 2024 · similarity 0.88)
- Going Beyond Multi-Task Dense Prediction with Synergy Embedding Models (CVPR 2024 · similarity 0.87)
- Swiss Army Knife: Synergizing Biases in Knowledge from Vision Foundation Models for Multi-Task Learning (ICLR 2025 · similarity 0.87)
- Task Preference Optimization: Improving Multimodal Large Language Models with Vision Task Alignment (CVPR 2025 · similarity 0.86)
- TaskGalaxy: Scaling Multi-modal Instruction Fine-tuning with Tens of Thousands Vision Task Types (ICLR 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.
