Research map / Vision-language and multimodal / Human and embodied modeling
How to Take a Memorable Picture? Empowering Users with Actionable Feedback
Francesco Laiti, Davide Talon, Jacopo Staiano, 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 memorability steering vector (MemCoach) to InternVL3.5 | InternVL3.5 8B · MemBench test | Improvement Ratio (IR): 0.73 -> 0.80 | Improved |
| add memorability steering vector to InternVL3.5 | InternVL3.5 8B · MemBench test | Relative Memorability RM (%): 5.47 -> 7.21 | Improved |
| add steering to InternVL3.5 | InternVL3.5 8B · MemBench test | Perplexity: 5.49 -> 4.99 | Improved |
| add steering to LLaVA-OV | LLaVA-OV 7B · MemBench test | IR: 0.70 -> 0.73 | Improved |
| add steering to Idefics3 | Idefics3 8B · MemBench test | IR / RM / perplexity: IR 0.73 -> 0.75; RM 6.64 -> 6.69 | No clear effect |
| add steering to Qwen2.5-VL | Qwen2.5-VL 7B · MemBench test | IR: 0.68 -> 0.74 | Improved |
| per-sample subtraction before averaging instead of averaging before differencing (ablation) | MemCoach steering vector · MemBench test | RM (%): 6.64 -> 7.21 | Improved |
| use InternVL3.5 as teacher instead of Qwen2.5-VL for contrasting data (ablation) | MemCoach · MemBench test | IR: 0.73 -> 0.80 | Improved |
| use steering instead of LoRA fine-tuning in low-data regime (ablation) | MemCoach · MemBench test | IR | Improved |
| prompt zero-shot MLLMs to rate memorability via yes/no likelihood | MLLMs · LaMem | Spearman: -0.06 to 0.08 vs 0.68 inter-annotator | No clear effect |
Closest papers by meaning
- Captured by Captions: On Memorization and its Mitigation in CLIP Models (ICLR 2025 · similarity 0.86)
- Exploring Local Memorization in Diffusion Models via Bright Ending Attention (ICLR 2025 · similarity 0.85)
- Déjà Vu Memorization in Vision–Language Models (NeurIPS 2024 · similarity 0.85)
- Re-evaluating Continual Learning with Few-Shot Adaptation (NeurIPS 2026 · similarity 0.85)
- VisMem: Latent Vision Memory Unlocks Potential of Vision-Language Models (CVPR 2026 · similarity 0.85)
- Detecting, Explaining, and Mitigating Memorization in Diffusion Models (ICLR 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.
