Research map / Architectures and efficiency / Neural architecture search
UFO: Unified Feature Optimization
Teng Xi, Yifan Sun, Deli Yu, Bi Li, Nan Peng, Gang Zhang, Xinyu Zhang, Zhigang Wang, Jinwen Chen, Jian Wang, Lufei Liu, Haocheng Feng and 4 more
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 task-specific FFN paths to the shared supernet (ablation) | UFO-Person · Market1501 | benchmark score (%): 87.62 -> 89.44 | Improved |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-Person · MSMT17 | benchmark score (%): 61.38 -> 64.84 | Improved |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-Person · Veri-776 | benchmark score (%): 86.50 -> 88.03 | Improved |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-Person · VehicleID | benchmark score (%): 85.46 -> 87.00 | Improved |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-Person · VeriWild | benchmark score (%): 67.92 -> 69.72 | Improved |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-Person · SOP | benchmark score (%): 83.50 -> 86.19 | Improved |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-SOP · CALFW | benchmark score (%): 95.92 -> 95.90 | No clear effect |
| add task-specific FFN paths to the shared supernet (ablation) | UFO-SOP · CPLFW | benchmark score (%): 93.37 -> 93.30 | Hurt |
| use FFN-path UFO-Person instead of all-shared weights | UFO-Person vs all shared · Market1501 | benchmark score (%): 88.19 -> 89.44 | Improved |
| use UFO for VehicleID instead of previous SOTA | UFO sub-network · VehicleID | benchmark score (%): 80.50 -> 87.04 | Improved |
| use UFO for VeriWild instead of previous SOTA | UFO sub-network · VeriWild | benchmark score (%): 77.30 -> 69.82 | Hurt |
| use UFO instead of Switch Transformers on VeriWild | UFO vs Switch Transformers · VeriWild | benchmark score (%): 12.32% | Improved |
Closest papers by meaning
- UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface (NeurIPS 2025 · similarity 0.83)
- Transferring Knowledge From Large Foundation Models to Small Downstream Models (ICML 2024 · similarity 0.79)
- Argus: A Compact and Versatile Foundation Model for Vision (CVPR 2025 · similarity 0.79)
- Improving Generalization of Universal Adversarial Perturbation via Dynamic Maximin Optimization (AAAI 2025 · similarity 0.79)
- Aux-NAS: Exploiting Auxiliary Labels with Negligibly Extra Inference Cost (ICLR 2024 · similarity 0.78)
- ALIEN: Implicit Neural Representations for Human Motion Prediction under Arbitrary Latency (CVPR 2025 · similarity 0.78)
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.
