Research map / Large language models / Language, safety and interpretability
Image is All You Need: Towards Efficient and Effective Large Language Model-Based Recommender Systems
Kibum Kim, Sein Kim, HongSeok Kang, Jiwan Kim, Heewoong Noh, Yeonjun In, Kanghoon Yoon, Jinoh Oh, Julian McAuley, Chanyoung Park
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 |
|---|---|---|---|
| add RISA alignment module to image-only features (ablation) | I-LLMRec · Amazon Sport | Hit@5: 0.3953 -> 0.4316 | Improved |
| add RISA alignment module to image-only features (ablation) | I-LLMRec · Amazon Art | Hit@5: 0.5040 -> 0.5564 | Improved |
| add CF feature to image in RERI (ablation) | I-LLMRec · Amazon Sport | Hit@5: 0.4316 -> 0.4491 | Improved |
| add text feature to image and CF in RERI (ablation) | I-LLMRec · Amazon Sport | Hit@5: 0.4491 -> 0.4570 | Improved |
| add text feature to image and CF in RERI (ablation) | I-LLMRec · Amazon Art | Hit@5: 0.5795 -> 0.5883 | Improved |
| add item descriptions to image tokens (I-LLMRec+D) | I-LLMRec · Amazon Sport | Hit@5: 0.4570 -> 0.4554 | No clear effect |
| represent items by images instead of summarized descriptions (I-LLMRec vs TRSR) | TRSR · Amazon Sport | Hit@5: 0.4302 -> 0.4570 | Improved |
| use I-LLMRec image-based alignment instead of UniMP multimodal LLM | UniMP · Amazon Phone | Hit@5: 0.4427 -> 0.5156 | Improved |
| fine-tune LLM backbone with LoRA instead of freezing | I-LLMRec · Amazon Sport | Hit@5: 0.4570 -> 0.4633 | Improved |
| replace long descriptions with a single image token per item | TRSR · inference time on long histories | inference time: 47 minutes -> 20 minutes including feature extraction | Improved |
| use I-LLMRec in a visually weak domain instead of description-based TRSR | TRSR · Goodbooks | Hit@5: 0.5100 -> 0.4870 | Hurt |
| use I-LLMRec instead of TALLRec on visual-centric data | TALLRec · Amazon Clothing | Hit@5: 0.4388 -> 0.5469 | Improved |
Closest papers by meaning
- Improving LLMs for Recommendation with Out-Of-Vocabulary Tokens (ICML 2025 · similarity 0.91)
- Can LLMs Outshine Conventional Recommenders? A Comparative Evaluation (NeurIPS 2025 · similarity 0.91)
- LLM-ESR: Large Language Models Enhancement for Long-tailed Sequential Recommendation (NeurIPS 2024 · similarity 0.89)
- Language Representations Can be What Recommenders Need: Findings and Potentials (ICLR 2025 · similarity 0.88)
- FACE: A General Framework for Mapping Collaborative Filtering Embeddings into LLM Tokens (NeurIPS 2025 · similarity 0.88)
- DOGE: LLMs-Enhanced Hyper-Knowledge Graph Recommender for Multimodal Recommendation (AAAI 2025 · similarity 0.88)
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.
