Research map / Large language models / Alignment and preferences
Towards Hands-Free Visual Dialog Interactive Recommendation
Tong Yu, Yilin Shen, Hongxia Jin
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 multimodal natural language feedback (SPR bandit) instead of click-only cascade bandits | interactive recommender · footwear, simulated, n=10 | average cumulative reward: about 3 times reward | Improved |
| use multimodal natural language feedback instead of click-only CascadeLinTS in real user study | interactive recommender · user study, K=4 | desired items found in 30 rounds: 4.7 -> 11.8 | Improved |
| use multimodal system instead of click-only CascadeLinTS in user study | interactive recommender · user study, K=4 | rounds to first desired item: 13 -> 6 | Improved |
| add visual dialog encoder output as context (C-CascadeLinTS) (ablation) | CascadeLinTS · footwear, simulated | average cumulative reward | No clear effect |
| use parameterized CascadeLinTS instead of CascadeUCB1 | cascade bandit · footwear, simulated | average cumulative reward | Improved |
| use SPR bandit exploration instead of greedy and epsilon-greedy with same multimodal inputs (ablation) | recommender search strategy · footwear, simulated | average cumulative reward | Improved |
| add epsilon-greedy random exploration (epsilon=0.1) to greedy (ablation) | greedy recommender · footwear, simulated | average cumulative reward | No clear effect |
| add pairwise logistic loss (ablation) | SPR bandit · footwear, simulated, including hard rare-item task | average cumulative reward | Improved |
| inject 25% or 50% noisy comments (ablation) | SPR bandit · footwear, simulated | average cumulative reward | Mixed |
Closest papers by meaning
- VisualLens: Personalization through Task-Agnostic Visual History (NeurIPS 2025 · similarity 0.87)
- Harnessing Multimodal Large Language Models for Multimodal Sequential Recommendation (AAAI 2025 · similarity 0.87)
- DOGE: LLMs-Enhanced Hyper-Knowledge Graph Recommender for Multimodal Recommendation (AAAI 2025 · similarity 0.86)
- MENTOR: Multi-level Self-supervised Learning for Multimodal Recommendation (AAAI 2025 · similarity 0.85)
- Modality-Independent Graph Neural Networks with Global Transformers for Multimodal Recommendation (AAAI 2025 · similarity 0.85)
- LGMRec: Local and Global Graph Learning for Multimodal Recommendation (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.
