Research map / Vision-language and multimodal / Human and embodied modeling
Adversarial Semantic Alignment for Improved Image Captions
Check 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 CIDEr regularization to the SCST GAN reward (ablation) | SCST Co-att GAN · COCO Test Set | CIDEr: 97.5 -> 111.1 | Helped |
| Add CIDEr regularization to the SCST GAN reward (ablation) | SCST Co-att GAN · COCO Test Set | Semantic Score: 0.190 -> 0.192 | Helped |
| Add CIDEr regularization to the SCST GAN reward (ablation) | SCST Co-att GAN · COCO Test Set | Vocabulary Coverage: 11.0 -> 7.3 | Hurt |
| Replace the co-attention discriminator with Joint-Emb (ablation) | SCST GAN · COCO Test Set | Semantic Score: 0.190 -> 0.188 | Hurt |
| Replace SCST with Gumbel ST (ablation) | Co-att GAN with log(D) reward · COCO Test Set | CIDEr: 97.5 -> 95.4 | Hurt |
| Add feature matching to Gumbel ST training (ablation) | Co-att GAN with log(D) reward · COCO Test Set | Semantic Score: 0.184 -> 0.175 | Hurt |
| Feed the previous attention context to the LSTM (ablation) | Captioner on COCO validation set · COCO validation set | CIDEr: 99.7 -> 103.3 | Helped |
| Use non-attentional captioners instead of attentive captioners (ablation) | Captioning models · OOC (Out of Context) | Semantic Score: 0.124 -> 0.111 | Hurt |
| Train with CIDEr-RL instead of SCST Co-att GAN | Captioner · COCO Test Set | Vocabulary Coverage: 9.2 -> 5.1 | Hurt |
Closest papers by meaning
- Text as Partial Constraint: Core–Residual Alignment for Robust Vision–Language Learning (NeurIPS 2026 · similarity 0.88)
- CapRL: Stimulating Dense Image Caption Capabilities via Reinforcement Learning (ICLR 2026 · similarity 0.87)
- Mining Fine-Grained Image-Text Alignment for Zero-Shot Captioning via Text-Only Training (AAAI 2024 · similarity 0.87)
- Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning (ICLR 2025 · similarity 0.87)
- Noise-Aware Image Captioning with Progressively Exploring Mismatched Words (AAAI 2024 · similarity 0.87)
- BACON: Improving Clarity of Image Captions via Bag-of-Concept Graphs (CVPR 2025 · similarity 0.87)
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.