Research map / Vision-language and multimodal / Human and embodied modeling
Towards Unique and Informative Captioning of Images
Zeyu Wang, Berthy Feng, Karthik Narasimhan, Olga Russakovsky
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 interpolated LM mutual-information re-ranking | TopDown · COCO test | SPICE-U: 23.70 -> 25.27 | Improved |
| add interpolated LM mutual-information re-ranking | TopDown · COCO test | GeoMean: 12.63 -> 12.94 | Improved |
| add interpolated LM mutual-information re-ranking | DiscCap · COCO test | SPICE-U: 23.74 -> 24.52 | Improved |
| add interpolated LM mutual-information re-ranking | AoANet · COCO test | SPICE-U: 26.04 -> 26.53 | Improved |
| add interpolated LM mutual-information re-ranking | TopDown · COCO test | CHAIRs: 8.68 -> 7.80 | Improved |
| add interpolated LM mutual-information re-ranking | TopDown · COCO test | CIDEr: 108.13 -> 106.42 | Hurt |
| add interpolated LM mutual-information re-ranking | DiscCap · COCO test | CIDEr: 112.39 -> 110.90 | Hurt |
| add interpolated LM mutual-information re-ranking | AoANet · COCO test | BLEU: 27.53 -> 27.18 | Hurt |
| use unigram LM instead of LSTM LM for re-ranking (ablation) | TopDown · COCO test | SPICE-U: 24.46 -> 25.05 | Improved |
| use LSTM LM instead of interpolated LM for re-ranking (ablation) | AoANet · COCO test | SPICE-U: 26.12 -> 26.53 | Improved |
| use LSTM LM only for re-ranking (ablation) | DiscCap · COCO test | SPICE-U: 23.74 -> 23.79 | No clear effect |
| use SPICE-U instead of SPICE as metric | PASCAL-50S · PASCAL-50S | Pearson correlation: 0.749 -> 0.767 | Improved |
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.
- SPICE — SPICE: Semantic Propositional Image Caption Evaluation (ECCV 2016) · 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
- SPICE: Semantic Propositional Image Caption Evaluation (ECCV 2016 · similarity 0.90)
- Improved Image Captioning via Policy Gradient Optimization of SPIDEr (ICCV 2017 · similarity 0.89)
- Guiding Image Captioning Models Toward More Specific Captions (ICCV 2023 · similarity 0.89)
- CAPability: A Comprehensive Visual Caption Benchmark for Evaluating Both Correctness and Thoroughness (NeurIPS 2025 · similarity 0.88)
- CaptionQA: Is Your Caption as Useful as the Image Itself? (CVPR 2026 · similarity 0.87)
- Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning (ICLR 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.
