Research map / Architectures and efficiency / Attention and transformers
Deep Attentive Ranking Networks for Learning to Order Sentences
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 |
|---|---|---|---|
| Use RankTxNet with pairwise or listwise ranking loss instead of pointwise regression | RankTxNet · Sentence ordering benchmarks | Kendall's tau | Helped |
| Use ListMLE loss instead of pointwise regression | RankTxNet · AAN abstracts | Kendall's tau: 0.7472 -> 0.7748 | Helped |
| Use ListMLE loss instead of pointwise regression | RankTxNet · SIND captions | Kendall's tau: 0.5510 -> 0.5652 | Helped |
| Use ListMLE loss instead of pointwise regression | RankTxNet · arXiv abstracts | PMR: 39.85 -> 43.44 | Helped |
| Use ListMLE loss instead of pointwise regression | RankTxNet · SIND captions | PMR: 14.03 -> 15.48 | Helped |
| Use ListMLE loss instead of pointwise regression | RankTxNet · ROCStory | PMR: 30.59 -> 38.02 | Mixed |
| Fine-tune BERT end-to-end on sentence ordering instead of freezing its parameters (ablation) | RankTxNet · AAN abstracts and SIND captions | Kendall's tau | Helped |
| Initialize the sentence encoder with pre-trained BERT instead of a randomly initialized Transformer (ablation) | RankTxNet · AAN abstracts and SIND captions | Kendall's tau | Helped |
| Use RankTxNet instead of ATTOrderNet | Order discrimination · Accidents | accuracy: 96.2 -> 96.79 | Helped |
| Use RankTxNet instead of ATTOrderNet | Order discrimination · Earthquakes | accuracy: 99.8 -> 98.65 | Hurt |
| Use RankTxNet ListMLE instead of ATTOrderNet | Sentence ordering · arXiv abstracts | PMR: 44.55 -> 43.44 | Hurt |
| Use RankTxNet ListMLE instead of HierarchicalATTNet | Sentence ordering · ROCStory | PMR: 39.62 -> 38.02 | Hurt |
Closest papers by meaning
- Sentence Ordering and Coherence Modeling using Recurrent Neural Networks (AAAI 2018 · similarity 0.88)
- Enhancing Pointer Network for Sentence Ordering with Pairwise Ordering Predictions (AAAI 2020 · similarity 0.87)
- Hierarchical Attention Networks for Sentence Ordering (AAAI 2019 · similarity 0.86)
- Contrastive Retrieval Heads for Improved Attention-Based Reranking (NeurIPS 2026 · similarity 0.84)
- Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankers (ICLR 2025 · similarity 0.83)
- SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking (NeurIPS 2026 · similarity 0.83)
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.