Research map / Probabilistic and scientific modeling / Bayesian optimization
MLink: Linking Black-Box Models From Multiple Domains for Collaborative Inference
Mu Yuan, Lan Zhang, Zimu Zheng, Yi-Nan Zhang, Xiang‐Yang Li
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 online periodic updating of model links (ablation) | MLink · Traffic monitoring vehicle-to-person | accuracy (%): 6.3 -> 70.2 | Improved |
| use loss-prediction-based active sampling instead of periodic (ablation) | MLink · Traffic monitoring vehicle-to-person | accuracy (%): 70.2 -> 74.7 | Improved |
| add uncertainty-based active sampling instead of periodic (ablation) | MLink · Traffic monitoring vehicle-to-person | accuracy (%): +3.3 | Improved |
| ensemble multi-source links instead of best single source (ablation) | MLink · Hollywood2 Action target | mAP (%): 29.7 -> 30.8 | Improved |
| ensemble multi-source links instead of best single source (ablation) | MLink · Hollywood2 Person target | IoU (%): 58.5 -> 59.2 | Improved |
| use model link instead of fine-tuning classifier | MLink · Smart building Gym | accuracy (%): 66.1 -> 88.1 | Improved |
| add global aggregated link from other domains (ablation) | MLink · Smart building Office | accuracy (%): 77.8 -> 90.2 | Improved |
| fuse local and global model links (ablation) | MLink · Smart building Gym | accuracy (%): up to 6% | Improved |
| fuse links from three cross-task models (ablation) | MLink · Smart building Action target | accuracy (%): up to 4.1% | Improved |
| use MLink instead of Reducto frame filtering (5GB) | MLink · City traffic | output accuracy (%): 84.1 -> 94 | Improved |
| use MLink instead of MTL (5GB) | MLink · Smart building | output accuracy (%): 53.3 -> 94.1 | Improved |
| apply model links to cross-domain semantic segmentation | MLink · Cityscapes to GTAV segmentation | accuracy: around 10% | Hurt |
Closest papers by meaning
- Knowledge Transfer from Vision Foundation Models for Efficient Training of Small Task-specific Models (ICML 2024 · similarity 0.83)
- Cloud-Device Collaborative Learning for Multimodal Large Language Models (CVPR 2024 · similarity 0.82)
- LEMON: Lossless model expansion (ICLR 2024 · similarity 0.82)
- msf-CNN: Patch-based Multi-Stage Fusion with Convolutional Neural Networks for TinyML (NeurIPS 2025 · similarity 0.82)
- DeepSpeed Data Efficiency: Improving Deep Learning Model Quality and Training Efficiency via Efficient Data Sampling and Routing (AAAI 2024 · similarity 0.82)
- Pfeife: Automatic Pipeline Parallelism for PyTorch (ICML 2025 · similarity 0.81)
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.
