Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Efficient Explorative Key-Term Selection Strategies for Conversational Contextual Bandits
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 joint estimation from arm-level and key-term-level feedback instead of separate estimates | ConLinUCB-UCB · Synthetic data, fixed key-term set | cumulative regret: 34.91% improvement | Helped |
| Select exploratory key-terms uniformly from a barycentric spanner | ConLinUCB-BS · Synthetic data, fixed key-term set | cumulative regret: 37.00% improvement over ConUCB | Helped |
| Select exploratory key-terms adaptively by maximum confidence radius | ConLinUCB-MCR · Synthetic data, fixed key-term set | cumulative regret: 43.10% improvement over ConUCB | Helped |
| Increase the conversation frequency | ConLinUCB-BS and ConLinUCB-MCR · Synthetic data | cumulative regret | Helped |
| Use barycentric-spanner key-term selection rather than ConUCB selection | ConLinUCB-BS · Synthetic data, |A_t|=25 | cumulative regret: 34.99% improvement over ConUCB | Helped |
| Use maximum-confidence-radius key-term selection rather than ConUCB selection | ConLinUCB-MCR · Synthetic data, |A_t|=25 | cumulative regret: 40.21% improvement over ConUCB | Helped |
| Use barycentric-spanner key-term selection rather than ConUCB selection | ConLinUCB-BS · Synthetic data, |A_t|=500 | cumulative regret: 50.36% improvement over ConUCB | Helped |
| Use maximum-confidence-radius key-term selection rather than ConUCB selection | ConLinUCB-MCR · Synthetic data, |A_t|=500 | cumulative regret: 53.77% improvement over ConUCB | Helped |
| Select available key-terms by maximum confidence radius | ConLinUCB-MCR · Synthetic data, time-varying key-term set | cumulative regret: 43.02% improvement over ConUCB | Helped |
| Use barycentric-spanner key-term selection | ConLinUCB-BS · Last.FM, fixed key-term set | cumulative regret: 13.28% improvement over ConUCB | Helped |
| Use maximum-confidence-radius key-term selection | ConLinUCB-MCR · Last.FM, fixed key-term set | cumulative regret: 17.12% improvement over ConUCB | Helped |
| Use barycentric-spanner key-term selection | ConLinUCB-BS · Movielens, fixed key-term set | cumulative regret: 13.08% improvement over ConUCB | Helped |
Closest papers by meaning
- A Multi-Agent Conversational Bandit Approach to Online Evaluation and Selection of User-Aligned LLM Responses (AAAI 2026 · similarity 0.88)
- Bandits with Preference Feedback: A Stackelberg Game Perspective (NeurIPS 2024 · similarity 0.86)
- Active Context Selection Improves Simple Regret in Contextual Bandits (NeurIPS 2026 · similarity 0.86)
- Pragmatic Feature Preferences: Learning Reward-Relevant Preferences from Human Input (ICML 2024 · similarity 0.85)
- MultiScale Contextual Bandits for Long Term Objectives (NeurIPS 2025 · similarity 0.85)
- ConvRec-R1: Training LLM-based Conversational Recommender Systems with Reinforcement Learning (ICLR 2026 · 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.