Research map / Vision-language and multimodal / CLIP and zero-shot
Stack-Captioning: Coarse-to-Fine Learning for Image Captioning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 stacked attention that progressively refines visual attention (ablation) | Stack-Cap · MSCOCO Karpathy test split | CIDEr: 115.9 -> 120.4 | Helped |
| Add coarse-to-fine decoders with intermediate supervision rather than a one-stage LSTM | Image captioning · MSCOCO Karpathy test split | CIDEr: 100.2 -> 115.9 | Helped |
| Add stacked attention to the coarse-to-fine model (ablation) | Stack-Cap · MSCOCO Karpathy test split | BLEU-4: 35.0 -> 36.1 | Helped |
| Add stacked attention to the coarse-to-fine model (ablation) | Stack-Cap · MSCOCO Karpathy test split | METEOR: 26.9 -> 27.4 | Helped |
| Add stacked attention to the coarse-to-fine model (ablation) | Stack-Cap · MSCOCO Karpathy test split | BLEU-1: 77.9 -> 78.6 | Helped |
| Add stacked attention to the coarse-to-fine model (ablation) | Stack-Cap · MSCOCO Karpathy test split | BLEU-2: 61.6 -> 62.5 | Helped |
| Add stacked attention to the coarse-to-fine model (ablation) | Stack-Cap · MSCOCO Karpathy test split | BLEU-3: 46.7 -> 47.9 | Helped |
| Add a third LSTM layer without intermediate supervision | LSTM image captioner · MSCOCO Karpathy test split | CIDEr: 91.4 -> 85.7 | Hurt |
| Use coarse-to-fine multi-stage learning instead of one-stage SCST | Image captioning · MSCOCO Karpathy test split | CIDEr: 100.2 -> 120.4 | Helped |
| Use a single model rather than a four-model ensemble | MSCOCO online evaluation · MSCOCO test server c40 | CIDEr: 116.7 -> 118.3 | Helped |
Closest papers by meaning
- CapRL: Stimulating Dense Image Caption Capabilities via Reinforcement Learning (ICLR 2026 · similarity 0.88)
- Temporal-Difference Learning With Sampling Baseline for Image Captioning (AAAI 2018 · similarity 0.88)
- Self-Critical N-Step Training for Image Captioning (CVPR 2019 · similarity 0.87)
- BalCapRL : A Balanced Framework for RL-Based MLLM Image Captioning (NeurIPS 2026 · similarity 0.87)
- Modeling Caption Diversity in Contrastive Vision-Language Pretraining (ICML 2024 · similarity 0.85)
- RePIC: Reinforced Post-Training for Personalizing Multi-Modal Language Models (NeurIPS 2025 · 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.