Research map / Learning theory and algorithms / Submodular and game algorithms
Auxiliary Learning as an Asymmetric Bargaining Game
Aviv Shamsian, Aviv Navon, Neta Glazer, Kenji Kawaguchi, Gal Chechik, Ethan Fetaya
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 |
|---|---|---|---|
| use asymmetric bargaining with dynamically learned task preferences | AuxiNash · NYUv2 | mIoU (semantic segmentation): 38.30 -> 40.79 | Improved |
| use asymmetric bargaining with dynamically learned task preferences | AuxiNash · Cityscapes | mIoU (semantic segmentation): 48.64 -> 52.52 | Improved |
| use asymmetric bargaining with dynamically learned task preferences | AuxiNash · CIFAR10-SSL-5K | test accuracy: 83.17 -> 87.01 | Improved |
| use asymmetric bargaining with dynamically learned task preferences | AuxiNash · Speech Commands SC-1000 | test accuracy: 97.0 ± 0.1 -> 97.2 ± 0.1 | Improved |
| optimize preferences using a 10% auxiliary set rather than the entire training set (ablation) | AuxiNash · Cityscapes | mIoU: 52.52 -> 51.81 | Hurt |
| optimize preferences using a 10% auxiliary set rather than the entire training set (ablation) | AuxiNash · NYUv2 | mIoU: 40.79 -> 38.94 | Hurt |
| fix the main-task preference at 0.7 rather than learn it dynamically (ablation) | AuxiNash · NYUv2 | mIoU: 40.79 -> 41.36 | Improved |
| fix the main-task preference at 0.8 rather than learn it dynamically (ablation) | AuxiNash · NYUv2 | mIoU: 40.79 -> 41.27 | Improved |
| fix the main-task preference at 0.9 rather than learn it dynamically (ablation) | AuxiNash · NYUv2 | mIoU: 40.79 -> 41.35 | Improved |
| change outer preference learning rate from 5e-3 to 1e-2 (ablation) | AuxiNash · NYUv2 | mIoU: 40.79 -> 40.75 | No clear effect |
| change outer preference learning rate from 5e-3 to 1e-3 (ablation) | AuxiNash · NYUv2 | mIoU: 40.79 -> 41.03 | No clear effect |
Closest papers by meaning
- AANG : Automating Auxiliary Learning (ICLR 2023 · similarity 0.85)
- Aux-NAS: Exploiting Auxiliary Labels with Negligibly Extra Inference Cost (ICLR 2024 · similarity 0.83)
- Fairness-Aware Meta-Learning via Nash Bargaining (NeurIPS 2024 · similarity 0.83)
- Cooperative Bargaining Games Without Utilities: Mediated Solutions from Direction Oracles (NeurIPS 2025 · similarity 0.82)
- NfgTransformer: Equivariant Representation Learning for Normal-form Games (ICLR 2024 · similarity 0.81)
- Task Groupings Regularization: Data-Free Meta-Learning with Heterogeneous Pre-trained Models (ICML 2024 · 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.
