Research map / Learning theory and algorithms / Prediction and decision losses
Preference Based Adaptation for Learning Objectives
Yao-Xiang Ding, Zhi-Hua Zhou
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 |
|---|---|---|---|
| adapt the objective weights with DL2M instead of fixed LIMO weights (w1=w2=1) (ablation) | LIMO · six multi-label datasets | average rank: 3.67 -> 2.42 | Improved |
| adapt hypothesis with Adapt-Boost using half the training iterations (ablation) | LIMO · six multi-label datasets | average rank: 3.67 -> 2.83 | Improved |
| combine DL2M and Adapt-Boost (ablation) | LIMO · six multi-label datasets | average rank: 3.67 -> 1.58 | Improved |
| add Adapt-Boost on top of DL2M (ADAPT-both vs ADAPT-obj) (ablation) | ADAPT-obj · six multi-label datasets | average rank: 2.42 -> 1.58 | Improved |
| LIMO (both margins) versus label-margin only (ablation) | LIMO objective · six multi-label datasets | average rank: 4.21 -> 3.67 | Improved |
| LIMO (both margins) versus instance-margin only (ablation) | LIMO objective · six multi-label datasets | average rank: 5.71 -> 3.67 | Improved |
| ADAPT-both versus LIMO | LIMO · emotions | ranking loss: .299 ± .023 -> .254 ± .020 | Improved |
| ADAPT-both versus LIMO | LIMO · bibtex | average precision: .567 ± .007 -> .581 ± .006 | Improved |
| ADAPT-both versus LIMO | LIMO · CAL500 | ranking loss: .182 ± .004 -> .181 ± .004 | No clear effect |
| ADAPT-both versus LIMO | LIMO · medical | average precision: .869 ± .026 -> .866 ± .025 | No clear effect |
| ADAPT-both versus LIMO | LIMO · medical | one-error: .163 ± .028 -> .176 ± .036 | Hurt |
| ADAPT-both versus LIMO | LIMO · CAL500 | one-error: .131 ± .053 -> .107 ± .024 | Improved |
Closest papers by meaning
- Direct Preference-Based Evolutionary Multi-Objective Optimization with Dueling Bandits (NeurIPS 2024 · similarity 0.85)
- Multi-Objective Bayesian Optimization with Active Preference Learning (AAAI 2024 · similarity 0.84)
- Neural Dueling Bandits: Preference-Based Optimization with Human Feedback (ICLR 2025 · similarity 0.83)
- FERERO: A Flexible Framework for Preference-Guided Multi-Objective Learning (NeurIPS 2024 · similarity 0.83)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.83)
- DOPL: Direct Online Preference Learning for Restless Bandits with Preference Feedback (ICLR 2025 · similarity 0.82)
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.
