Research map / Vision-language and multimodal / Multimodal LLMs
ReflectCAP: Detailed Image Captioning with Reflective Memory
Kyungmin Min, Minbeom Kim, Kang-il Lee, Seunghyun Yoon, Kyomin Jung
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 Structured Reflection Notes (ReflectCAP) | InternVL3.5-4B zero-shot · IIW-400 | F1: 59.1 -> 63.0 | Improved |
| add Structured Reflection Notes (ReflectCAP) | Qwen3-VL-32B zero-shot · IIW-400 | F1: 75.7 -> 76.4 | Improved |
| add Structured Reflection Notes (ReflectCAP) | GPT-4.1-mini zero-shot · CapArena-Auto | score: 57.7 -> 90.0 | Improved |
| add few-shot human exemplars | InternVL3.5-4B · IIW-400 | F1: 59.1 -> 51.1 | Hurt |
| add CapMAS multi-agent verification | GPT-4.1-nano · CapArena-Auto | score: 21.2 -> -14.7 | Hurt |
| add self-correction | Qwen2.5-7B · IIW-400 | F1: 62.7 -> 60.4 | Hurt |
| add SFT on human-authored DOCCI captions | Qwen2.5-VL-7B · IIW-400 | F1: 62.7 -> 57.6 | Hurt |
| add SFT on ReflectCAP-generated captions | Qwen2.5-VL-7B · IIW-400 | F1: 62.7 -> 66.5 | Improved |
| add separate passes for Avoid and Include notes (vs combined prompt) (ablation) | ReflectCAP on InternVL3.5-4B · IIW-400 (100 images) | F1: 57.3 -> 64.4 | Improved |
| add separate passes for Avoid and Include notes (vs combined prompt) (ablation) | ReflectCAP on GPT-4.1-mini · IIW-400 (100 images) | F1: 77.1 -> 78.3 | Improved |
| add Grounded Base Caption (Avoid notes only) (ablation) | Qwen2.5-VL family · IIW-400 | factuality: marginal improvement or degradation | Mixed |
| increase exemplar images N from 30 to 100 (ablation) | ReflectCAP · IIW-400 (100 images) | F1: plateau around N=30, slight decline at N=100 | No clear effect |
Closest papers by meaning
- Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning (ICLR 2025 · similarity 0.90)
- Top-Down Semantic Refinement for Image Captioning (AAAI 2026 · similarity 0.90)
- MM-SeR: Multimodal Self-Refinement for Lightweight Image Captioning (CVPR 2026 · similarity 0.89)
- CaptionQA: Is Your Caption as Useful as the Image Itself? (CVPR 2026 · similarity 0.89)
- Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning (CVPR 2025 · similarity 0.88)
- CapsFusion: Rethinking Image-Text Data at Scale (CVPR 2024 · 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.
