Research map / Vision-language and multimodal / CLIP and zero-shot
Seeing What’s Not There: Negation Understanding Needs More Than Training
Bhuvan Aggarwal, Amit More, Mudit Soni, S Divakar Bhat
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 embedding correction (negated-concept projection removal) | CLIP · NegBench MCQ COCO | MCQ accuracy: 24.7 -> 72.5 | Improved |
| add embedding correction | CLIP · NegBench MCQ VOC2007 | MCQ accuracy: 24.3 -> 78.6 | Improved |
| add embedding correction | CLIP · NegBench Retrieval COCO | R-Neg@5: 57.3 -> 63.2 | Improved |
| add embedding correction | NegCLIP* · NegBench MCQ COCO | MCQ accuracy: 56.2 -> 69.5 | Improved |
| add embedding correction | NegCLIP* · NegBench Retrieval MSRvtt | R-Neg@5: 51.5 -> 52.2 | Improved |
| add embedding correction | SigLIP2 · NegBench MCQ COCO | MCQ accuracy: 27.2 -> 66.3 | Improved |
| add embedding correction | CLIP · VALSE Existence | accuracy: 71.0 -> 79.5 | Improved |
| add embedding correction | CLIP · Flickr30K negation subset | text retrieval Acc R@1: 34.0 -> 41.3 | Improved |
| add embedding correction | CLIP · NegBench MCQ COCO (negated captions) | confidence score: 0.16 -> 0.64 | Improved |
| use custom negation scope rules instead of off-the-shelf NegEx (ablation) | Ours + CLIP · NegBench MCQ COCO | MCQ accuracy: 61.3 -> 72.5 | Improved |
| use neutral anchor word instead of arbitrary noun (ablation) | Ours + CLIP · NegBench MCQ COCO | MCQ accuracy: 50.78 -> 72.51 | Improved |
| choose among neutral anchor words (neutral, balanced, unbiased, fair) (ablation) | Ours + CLIP · NegBench MCQ COCO | MCQ accuracy: 72.42 -> 72.54 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TripletCLIP — TripletCLIP: Improving Compositional Reasoning of CLIP via Synthetic Vision-Language Negatives (NeurIPS 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Vision-Language Models Do Not Understand Negation (CVPR 2025 · similarity 0.92)
- What "Not" to Detect: Negation-Aware VLMs via Structured Reasoning and Token Merging (ICLR 2026 · similarity 0.90)
- Interpreting and Analysing CLIP's Zero-Shot Image Classification via Mutual Knowledge (NeurIPS 2024 · similarity 0.87)
- Enhancing Compositional Reasoning in CLIP via Reconstruction and Alignment of Text Descriptions (NeurIPS 2025 · similarity 0.86)
- Building Vision-Language Models on Solid Foundations with Masked Distillation (CVPR 2024 · similarity 0.86)
- TripletCLIP: Improving Compositional Reasoning of CLIP via Synthetic Vision-Language Negatives (NeurIPS 2024 · similarity 0.86)
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.
