Research map / Architectures and efficiency / Attention and transformers
Syntactically Look-Ahead Attention Network for Sentence Compression
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 |
|---|---|---|---|
| Track parent and child dependency words with a selective gate | SLAHAN · Google dataset ALL | R-1: 78.5 -> 79.3 | Helped |
| Track parent and child dependency words with a selective gate | SLAHAN · BNC Corpus long sentences | F1: 55.4 -> 57.7 | Helped |
| Add child recursive attention (ablation) | SLAHAN · Google dataset LONG | F1: 82.8 -> 83.3 | Helped |
| Add parent recursive attention (ablation) | SLAHAN · Google dataset LONG | F1: 83.0 -> 83.3 | Helped |
| Add child recursive attention (ablation) | SLAHAN · BNC Corpus long sentences | F1: 54.2 -> 57.7 | Helped |
| Add parent recursive attention (ablation) | SLAHAN · BNC Corpus long sentences | F1: 55.6 -> 57.7 | Helped |
| Remove syntax-tree supervision (ablation) | SLAHAN · Google dataset ALL | F1: 85.5 -> 85.4 | No clear effect |
| Remove syntax-tree supervision (ablation) | SLAHAN · BNC Corpus long sentences | F1: 57.7 -> 54.6 | Hurt |
| Track only child words instead of only parent words (ablation) | SLAHAN ablations · Google dataset LONG | R-2: 67.0 -> 67.3 | Helped |
| Use only parent attention instead of the gated parent-child combination (ablation) | SLAHAN · Google dataset ALL | F1: 85.5 -> 85.0 | Hurt |
| Use only child attention instead of the gated parent-child combination (ablation) | SLAHAN · Google dataset ALL | F1: 85.5 -> 85.4 | No clear effect |
| Use only child attention instead of the gated parent-child combination (ablation) | SLAHAN · BNC Corpus long sentences | F1: 57.7 -> 55.6 | Hurt |
Closest papers by meaning
- Word Attention for Sequence to Sequence Text Understanding (AAAI 2018 · similarity 0.87)
- Prompt Compression with Context-Aware Sentence Encoding for Fast and Improved LLM Inference (AAAI 2025 · similarity 0.85)
- FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference (ICLR 2025 · similarity 0.85)
- MVA: Linear Attention with High-order Query-Keys Integration and Multi-level Vocabulary Decomposition (ICML 2025 · similarity 0.85)
- SALS: Sparse Attention in Latent Space for KV Cache Compression (NeurIPS 2025 · similarity 0.84)
- UniGist: Towards General and Hardware-aligned Sequence-level Long Context Compression (NeurIPS 2025 · similarity 0.84)
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.