Research map / Vision-language and multimodal / CLIP and zero-shot
Negative Entity Suppression for Zero-Shot Captioning with Synthetic Images
Zimao Lu, Hui Xu, Bing Liu, Ke Wang
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 attention-level suppression of negative-entity features (ablation) | NES · Flickr30k->COCO | CIDEr: 83.9 -> 87.5 | Improved |
| add attention-level suppression (ablation) | NES · Flickr30k->COCO | CHAIR-S: 11.8 -> 12.1 | Hurt |
| add attention-level suppression (ablation) | NES · Flickr30k->COCO | entity recall: 44.1 -> 44.9 | Improved |
| enable all four components (SIR, SIF, NEF, AS) vs none (ablation) | NES · Flickr30k->COCO | CIDEr: 76.3 -> 87.5 | Improved |
| enable all four components vs none (ablation) | NES · Flickr30k->COCO | CHAIR-S: 22.8 -> 12.1 | Improved |
| use synthetic-image retrieval instead of text retrieval (ablation) | NES · Flickr30k->COCO | hallucination rate: hallucinations decrease about 50% | Improved |
| lower suppression strength lambda below 0.3 (ablation) | NES · Flickr30k->COCO | CIDEr | Hurt |
| CLIPScore-based fusion vs fixed fusion weights (ablation) | NES · Flickr30k->COCO | CIDEr / CHAIR | Mixed |
| NES vs IFCap | NES · Flickr30k->COCO (no target corpus) | CIDEr: 60.7 -> 69.9 | Improved |
| NES vs IFCap | NES · Flickr30k test, in-domain | METEOR: 22.9 -> 22.1 | Hurt |
Closest papers by meaning
- Zero-Shot Image Captioning with Multi-type Entity Representations (AAAI 2025 · similarity 0.89)
- MeaCap: Memory-Augmented Zero-shot Image Captioning (CVPR 2024 · similarity 0.87)
- Mining Fine-Grained Image-Text Alignment for Zero-Shot Captioning via Text-Only Training (AAAI 2024 · similarity 0.86)
- Semantic Surgery: Zero-Shot Concept Erasure in Diffusion Models (NeurIPS 2025 · similarity 0.86)
- Noise-Aware Image Captioning with Progressively Exploring Mismatched Words (AAAI 2024 · similarity 0.85)
- Image Captioning with Multi-Context Synthetic Data (AAAI 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.
