Research map / Vision-language and multimodal / CLIP and zero-shot
Not Just What’s There: Enabling CLIP to Comprehend Negated Visual Descriptions Without Fine-Tuning
Junhao Xiao, Zhiyu Wu, Hao Lin, Yi Chen, Yahui Liu, Xiaoran Zhao, Zixu Wang, Zejiang He
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 syntax stream in Lens (ablation) | CLIPGlasses · CC-Neg | accuracy: 94.09% -> 96.56% | Improved |
| add syntax stream in Lens (ablation) | CLIPGlasses · CC-Neg | FAR: 4.4927 -> 7.8253 | Improved |
| add semantic stream in Lens (ablation) | CLIPGlasses · CC-Neg | accuracy: 94.93% -> 96.56% | Improved |
| add residual gating in Lens (ablation) | CLIPGlasses · CC-Neg | accuracy: 68.93% -> 96.56% | Improved |
| add cross-modal context in Frame (ablation) | CLIPGlasses · CC-Neg | accuracy: 93.79% -> 96.56% | Improved |
| add dynamic repulsion weight (Frame) (ablation) | CLIPGlasses · CC-Neg | accuracy: 63.74% -> 96.56% | Improved |
| add cross-modal context in Frame (ablation) | CLIPGlasses · CC-Neg | FAR: 7.3051 -> 7.8253 | Improved |
| CLIPGlasses vs CoN-CLIP trained on CC-Neg | CoN-CLIP · Neg-COCO-MCQ | accuracy: 25.70% -> 34.51% | Improved |
| CLIPGlasses vs CoN-CLIP trained on Neg-COCO-R | CoN-CLIP · CC-Neg-val | accuracy: 65.91% -> 93.36% | Improved |
| CLIPGlasses vs CoN-CLIP trained on Neg-COCO-R | CoN-CLIP · Neg-COCO-MCQ | accuracy: 30.61% -> 35.90% | Improved |
| CLIPGlasses vs CoN-CLIP in-domain | CoN-CLIP · CC-Neg-val | accuracy: 99.70% -> 96.56% | Hurt |
| fine-tune CLIP text encoder on negation data (CoN-CLIP) vs vanilla | Vanilla CLIP · ImageNet zero-shot | accuracy: 53.87 -> 50.98 | Hurt |
Closest papers by meaning
- VITRIX-CLIPIN: Enhancing Fine-Grained Visual Understanding in CLIP via Instruction-Editing Data and Long Captions (NeurIPS 2025 · similarity 0.89)
- Vision-Language Models Do Not Understand Negation (CVPR 2025 · similarity 0.89)
- Interpreting and Analysing CLIP's Zero-Shot Image Classification via Mutual Knowledge (NeurIPS 2024 · similarity 0.88)
- UniLiP: Adapting CLIP for Unified Multimodal Understanding, Generation and Editing (ICLR 2026 · similarity 0.88)
- Building Vision-Language Models on Solid Foundations with Masked Distillation (CVPR 2024 · similarity 0.88)
- Enhancing Compositional Reasoning in CLIP via Reconstruction and Alignment of Text Descriptions (NeurIPS 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.
