Research map / Large language models / Instruction and fine-tuning
Landmark-Guided Policy Optimization for Multi-Objective Language Model Selection
Marcio Monteiro, Weichen Li, Puyu Wang, Marius Kloft, Sophie Fellenz
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 learned RL allocation policy with landmark fine-tuning instead of MO-ASHA | LAMPS · 12 held-out datasets | time to 99% of optimal hypervolume: best on 9 of 12 datasets vs 3 for MO-ASHA | Improved |
| use LAMPS instead of exhaustive search | LLM selection for fine-tuning · Amazon (Cell Phone) | cost (USD): $5,141.67 -> $449.58 | Improved |
| use LAMPS instead of exhaustive search | LLM selection for fine-tuning · 12 held-out datasets | search time: 73.6% reduction | Improved |
| use LAMPS instead of MO-ASHA for stability | LAMPS · 12 held-out datasets | search time stability | Improved |
| LAMPS fails to reach optimal hypervolume promptly in some cases | LAMPS · 12 held-out datasets | time to optimality: near-optimal quickly in 9 of 12 cases | Mixed |
| use sparse reward |X\Xt|/delta_t instead of hypervolume minus lambda|X| (ablation) | PPO policy · policy training | training stability | Improved |
| add invalid action masking for completed models (ablation) | RL policy · policy training | exploration efficiency | Improved |
| meta-learn on 11 datasets before deployment | LAMPS · leave-one-out over 12 datasets | payback in future tasks: 5,268 GPU-hours amortized after 13.9 tasks | Improved |
Closest papers by meaning
- Enhancing Decision-Making of Large Language Models via Actor-Critic (ICML 2025 · similarity 0.88)
- MetaAligner: Towards Generalizable Multi-Objective Alignment of Language Models (NeurIPS 2024 · similarity 0.88)
- Model Fusion through Bayesian Optimization in Language Model Fine-Tuning (NeurIPS 2024 · similarity 0.87)
- metaTextGrad: Automatically optimizing language model optimizers (NeurIPS 2025 · similarity 0.87)
- ToolSearcher: Optimizing Tool Selection at Scale via Reinforcement Learning (NeurIPS 2026 · similarity 0.87)
- Unleashing the Potential of Large Language Models as Prompt Optimizers: Analogical Analysis with Gradient-based Model Optimizers (AAAI 2025 · similarity 0.87)
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.
