Research map / Vision-language and multimodal / Human and embodied modeling
Positive-Augmented Contrastive Learning for Image and Video Captioning Evaluation
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 |
|---|---|---|---|
| Fine-tune CLIP projections using curated COCO pairs plus generated image and caption positives | CLIP-S caption evaluation · Flickr8k-Expert | Kendall τ_b: 51.1 -> 53.9; +2.8 | Helped |
| Fine-tune CLIP projections using curated COCO pairs plus generated image and caption positives | CLIP-S caption evaluation · Flickr8k-CF | Kendall τ_b: 34.4 -> 36.0; +1.6 | Helped |
| Fine-tune CLIP projections using curated COCO pairs plus generated image and caption positives | CLIP-S caption evaluation · Composite | Kendall τ_b: 49.8 -> 51.5; +1.7 | Helped |
| Use positive-augmented contrastive embeddings for video-caption scoring | EMScore · VATEX-EVAL, No Ref | Spearman ρ: 30.3 -> 32.6; +2.3 | Helped |
| Use positive-augmented contrastive embeddings for video-caption scoring | EMScore · VATEX-EVAL, 1 Ref | Spearman ρ: 37.1 -> 40.5; +3.4 | Helped |
| Use positive-augmented contrastive embeddings for video-caption scoring | EMScore · VATEX-EVAL, 9 Refs | Spearman ρ: 47.2 -> 48.8; +1.6 | Helped |
| Use positive-augmented contrastive embeddings | CLIP-S · Pascal-50S | Mean accuracy: Mean: 80.9 -> 82.4; +1.5 | Mixed |
| Use positive-augmented contrastive embeddings | CLIP-S · Abstract-50S | Mean accuracy: 68.2 -> 69.7; +1.5 | Helped |
| Use positive-augmented contrastive embeddings | CLIP-S · FOIL | Accuracy: 87.2 -> 89.9; +2.7 | Helped |
| Use positive-augmented contrastive embeddings | EMScore · ActivityNet-FOIL | Accuracy: 89.5 -> 90.1; +0.6 | Helped |
| Include reference captions in positive-augmented evaluation | PAC-S · FOIL | Accuracy: 89.9 -> 93.7 | Helped |
| Use CLIP ViT-L/14 rather than CLIP ViT-B/32 features | PAC-S · Flickr8k-Expert | Kendall τ_b: 53.9 -> 55.1 | Helped |
Closest papers by meaning
- CLIPLoss and Norm-Based Data Selection Methods for Multimodal Contrastive Learning (NeurIPS 2024 · similarity 0.86)
- CLIP-KD: An Empirical Study of CLIP Model Distillation (CVPR 2024 · similarity 0.86)
- Retrieval-Enhanced Contrastive Vision-Text Models (ICLR 2024 · similarity 0.86)
- Post-pre-training for Modality Alignment in Vision-Language Foundation Models (CVPR 2025 · similarity 0.86)
- Modeling Caption Diversity in Contrastive Vision-Language Pretraining (ICML 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.