Research map / Vision-language and multimodal / Multimodal LLMs
Training-Free Personalization via Retrieval and Reasoning on Fingerprints
Deepayan Das, Davide Talon, Yiming Wang, Massimiliano Mancini, Elisa Ricci
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 fingerprint attributes to CoT reasoning (ablation) | R2P (CoT, no pairwise) · PerVA | captioning recall: 62.8 -> 71.2 | Improved |
| add fingerprint attributes to CoT reasoning (ablation) | R2P (CoT, no pairwise) · PerVA | weighted recognition: 84.7 -> 91.8 | Improved |
| add CoT reasoning alone, without fingerprint attributes (ablation) | R2P naive prompting · PerVA | weighted recognition: 86.5 -> 84.7 | Hurt |
| add pairwise reasoning on top of CoT and fingerprints (ablation) | R2P · PerVA | captioning recall: 71.2 -> 72.5 | Improved |
| add pairwise reasoning only (ablation) | R2P naive prompting · PerVA | weighted recognition: 86.5 -> 92.3 | Improved |
| add CoT to pairwise reasoning without attributes (ablation) | R2P · PerVA | recognition and captioning: 92.3 -> 91.6 Wtd, 65.9 -> 67.3 recall | Mixed |
| use human-specified attributes instead of VLM fingerprints (ablation) | R2P · PerVA | captioning recall: 72.5 -> 72.8 | No clear effect |
| use attribute verification instead of abstention (ablation) | R2P · PerVA | captioning recall: 70.7 -> 72.5 | Improved |
| use attribute verification instead of logits-based check (ablation) | R2P · PerVA | captioning recall: 70.9 -> 72.5 | Improved |
| use attribute verification instead of always running pairwise reasoning (ablation) | R2P · PerVA | captioning recall: 72.3 -> 72.5 | No clear effect |
| use attribute verification instead of no estimation (ablation) | R2P · Yo'LLaVA | captioning recall: 82.0 -> 91.5 | Improved |
| use fused image-text retrieval instead of CLIP image only (ablation) | R2P · PerVA | HIT@3: 76.9 -> 87.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- Yo'LLaVA — Yo'LLaVA: Your Personalized Language and Vision Assistant (NeurIPS 2024) · better on 4 of 4 shared results
- MyVLM — MyVLM: Personalizing VLMs for User-Specific Queries (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MM-R1: Unleashing the Power of Unified Multimodal Large Language Models for Personalized Image Generation (AAAI 2026 · similarity 0.88)
- MMPB: It’s Time for Multi-Modal Personalization (NeurIPS 2025 · similarity 0.88)
- Ego: Embedding-Guided Personalization of Vision-Language Models (CVPR 2026 · similarity 0.87)
- Unified Personalized Understanding, Generating and Editing (CVPR 2026 · similarity 0.87)
- Improving Personalized Search with Regularized Low-Rank Parameter Updates (CVPR 2025 · similarity 0.87)
- Contextualized Visual Personalization in Vision-Language Models (ICML 2026 · similarity 0.86)
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.
