Research map / Vision-language and multimodal / Human and embodied modeling
Prophet Attention: Predicting Attention with Future Attention
Fenglin Liu, Xuancheng Ren, Xian Wu, Shen Ge, Wei Fan, Yuexian Zou, Xu Sun
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 Dynamic Prophet Attention | AoANet · MSCOCO | CIDEr: 129.6 -> 133.4 | Improved |
| add Dynamic Prophet Attention | GVD · Flickr30k Entities | CIDEr: 59.9 -> 62.7 | Improved |
| add Dynamic Prophet Attention | GVD · Flickr30k Entities | F1loc: 11.8 -> 15.5 | Improved |
| add Dynamic Prophet Attention | Up-Down · MSCOCO | CIDEr: 123.5 -> 129.0 | Improved |
| use DPA (noun-phrase windows, masked non-visual words) instead of CPA (ablation) | Prophet Attention · MSCOCO | CIDEr: 126.3 -> 128.1 | Improved |
| mask Prophet Attention on non-visual words (ablation) | DPA · MSCOCO | CIDEr: 0.5 CIDEr | Improved |
| increase regularization lambda from 0.1 to 0.3 (ablation) | DPA · MSCOCO | CIDEr: 129.0 -> 121.3 | Hurt |
| increase lambda from 0.1 to 1 (ablation) | DPA · MSCOCO | CIDEr: 129.0 -> 114.8 | Hurt |
| lower lambda from 0.1 to 0.05 (ablation) | DPA · MSCOCO | CIDEr: 129.0 -> 128.1 | Hurt |
| use L1 attention loss instead of L2 (ablation) | DPA · MSCOCO | CIDEr: 128.2 -> 129.0 | Improved |
| use L1 attention loss instead of KL divergence (ablation) | DPA · MSCOCO | CIDEr: 126.9 -> 129.0 | Improved |
| add DPA to paraphrase model | LSTM-Attention · MSCOCO paraphrase | BLEU: 29.2 -> 36.5 | Improved |
Closest papers by meaning
- Posterior Attention Models for Sequence to Sequence Learning (ICLR 2019 · similarity 0.80)
- Attention Prompting on Image for Large Vision-Language Models (ECCV 2024 · similarity 0.79)
- Word Attention for Sequence to Sequence Text Understanding (AAAI 2018 · similarity 0.79)
- Hierarchical LSTMs with Adaptive Attention for Visual Captioning (TPAMI 2019 · similarity 0.79)
- Attention Alignment Between Humans and Vision-Language Models (NeurIPS 2026 · similarity 0.77)
- Unveiling and Harnessing Hidden Attention Sinks: Enhancing Large Language Models without Training through Attention Calibration (ICML 2024 · similarity 0.77)
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.
