Research map / Bandits and online learning / Multi-armed bandits
Efficient Multi-objective Prompt Optimization via Pure-exploration Bandits
Donghao Li, Chengshuai Shi, Weijuan Ou, Cong Shen, Jing Yang
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 |
|---|---|---|---|
| use bandit-based CSR instead of uniform evaluation | Gensec · XSum (Gemma-7B, K=100, b=10) | soft constrained reward: 0.044 -> 0.143 | Improved |
| use MLP-CSR (shared reward structure) instead of independent CSR | Gensec · XSum (Gemma-7B, K=100, b=10) | soft constrained reward: 0.143 -> 0.147 | Improved |
| use MLP-CSR instead of CSR at K=30, b=10 | Gensec · XSum (Gemma-7B, K=30, b=10) | soft constrained reward: 0.143 -> 0.142 | No clear effect |
| use EGE instead of uniform evaluation | GenPSI · CNN/DailyMail (Llama3-8B, K=100, b=10) | hypervolume: 0.1631 -> 0.1754 | Improved |
| use MLP-EGE instead of uniform evaluation | GenPSI · CNN/DailyMail (Llama3-8B, K=100, b=3) | hypervolume: 0.1519 -> 0.1628 | Improved |
| use MLP-EGE instead of EGE at K=50, b=10 | GenPSI · CNN/DailyMail (Llama3-8B, K=50, b=10) | hypervolume: 0.1720 -> 0.1673 | Hurt |
| increase per-arm budget b | Gensec and GenPSI · XSum / CNN/DailyMail | soft constrained reward | Improved |
| increase candidate set size K | Gensec · XSum / CNN/DailyMail | soft constrained reward | No clear effect |
Closest papers by meaning
- Efficient Prompt Optimization Through the Lens of Best Arm Identification (NeurIPS 2024 · similarity 0.92)
- MASPOB: Bandit-Based Prompt Optimization for Multi-Agent Systems with Graph Neural Networks (ICML 2026 · similarity 0.87)
- Sample Efficient Demonstration Selection for In-Context Learning (ICML 2025 · similarity 0.86)
- Bandits with Preference Feedback: A Stackelberg Game Perspective (NeurIPS 2024 · similarity 0.86)
- On Speeding Up Language Model Evaluation (ICLR 2025 · similarity 0.86)
- Multiobjective Lipschitz Bandits under Lexicographic Ordering (AAAI 2024 · similarity 0.85)
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.
