Research map / Learning theory and algorithms / Prediction and decision losses
Decomposition Strategies for Constructive Preference Elicitation
Paolo Dragone, Stefano Teso, Mohit Kumar, Andrea Passerini
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 part-wise PCL instead of complete-configuration Coactive Learning | Synthetic grid problem · Synthetic grid | complete-configuration regret: quality comparable to CL after enough queries are made | No clear effect |
| Use part-wise rather than complete-configuration inference | PCL · Synthetic grid | cumulative runtime: The runtime gap between full and part-wise inference is already clear | Improved |
| Select parts randomly rather than using the smallest-part-first strategy (ablation) | PCL · Synthetic grid | complete-configuration regret | Improved |
| Select parts randomly rather than using UCB1 (ablation) | PCL · Synthetic grid | complete-configuration regret | Improved |
| Use smallest-part-first selection rather than random selection (ablation) | PCL · Synthetic grid | complete-configuration regret | Hurt |
| Use UCB1 selection rather than random selection (ablation) | PCL · Synthetic grid | complete-configuration regret | Hurt |
| Increase user informativeness alpha (ablation) | PCL · Training planning | regret: from 0.1, to 0.3, to 0.5 | Improved |
| Use part-wise PCL | PCL · Training planning | regret: after about 50 iterations or less | Improved |
| Use alpha 0.3 rather than alpha 0.5 (ablation) | PCL · Hotel planning | regret: an alpha of 0.3 achieves basically the same results as those for alpha equal to 0.5 | No clear effect |
| Use part-wise PCL | PCL · Hotel planning | regret: by an order of magnitude in around a 100 iterations | Improved |
| Use informed part-selection strategies rather than random selection (ablation) | PCL · Constructive preference elicitation tasks | performance: do not yield a significant performance improvement | No clear effect |
Closest papers by meaning
- Learning from negative feedback, or positive feedback or both (ICLR 2025 · similarity 0.86)
- Comparing Few to Rank Many: Active Human Preference Learning Using Randomized Frank-Wolfe Method (ICML 2025 · similarity 0.86)
- Preference Learning of Latent Decision Utilities with a Human-like Model of Preferential Choice (NeurIPS 2024 · similarity 0.85)
- Discovering Preference Optimization Algorithms with and for Large Language Models (NeurIPS 2024 · similarity 0.85)
- Optimal Design for Human Preference Elicitation (NeurIPS 2024 · similarity 0.85)
- Explicit Preference Optimization: No Need for an Implicit Reward Model (ICML 2025 · 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.
