Research map / Graph learning / Knowledge and temporal graphs
Enhancing Scientific Papers Summarization with Citation Graph
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 citation-graph features with the source document (ablation) | CgSum · SSN (inductive) | R-1: 42.84 -> 44.36 | Helped |
| Use citation-graph features with the source document (ablation) | CgSum · SSN (inductive) | R-L: 37.59 -> 39.43 | Helped |
| Add neighborhood extraction to select informative citation subgraphs (ablation) | CgSum · SSN (inductive) | R-1: 44.23 -> 44.36 | Helped |
| Add residual connections to the graph encoder (ablation) | CgSum · SSN (inductive) | R-L: 38.95 -> 39.43 | Helped |
| Use trigram blocking during decoding (ablation) | CgSum · SSN (inductive) | R-1: 43.48 -> 44.36 | Helped |
| Add ROUGE credit from a relevant neighbor abstract during beam search (ablation) | CgSum · SSN (inductive) | R-L: 38.70 -> 39.43 | Helped |
| Add ROUGE credit from a relevant neighbor abstract during beam search (ablation) | CgSum · SSN (inductive) | R-2: 14.49 -> 14.69 | Helped |
| Use 1-hop rather than 2-hop neighbors | CgSum · SSN (inductive) | R-1: 44.36 vs 44.28 | Mixed |
| Use 1-hop rather than 2-hop neighbors | CgSum · SSN (transductive) | R-L: 38.89 -> 39.10 | Helped |
| Concatenate neighbors' abstracts to the source document instead of encoding graph structure | PTGen + Cov · SSN (inductive) | R-1: 42.84 -> 43.05 | No clear effect |
| Add copy mechanism | Transformer Abs · SSN (inductive) | R-1: 37.78 -> 43.35 | Helped |
| Extend BERT position embeddings to 640 | BertSumAbs · SSN (inductive) | R-1: 41.22 -> 43.73 | Helped |
Closest papers by meaning
- ScisummNet: A Large Annotated Corpus and Content-Impact Models for Scientific Paper Summarization with Citation Networks (AAAI 2019 · similarity 0.91)
- HLM-Cite: Hybrid Language Model Workflow for Text-based Scientific Citation Prediction (NeurIPS 2024 · similarity 0.87)
- DisenCite: Graph-Based Disentangled Representation Learning for Context-Specific Citation Generation (AAAI 2022 · similarity 0.85)
- COMPOSE: Composing Future Theorems from Citations and Formal Structure (NeurIPS 2026 · similarity 0.81)
- Language Models of Code Are Few-Shot Planners and Reasoners for Multi-Document Summarization with Attribution (AAAI 2025 · similarity 0.81)
- Mixture of Knowledge Minigraph Agents for Literature Review Generation (AAAI 2025 · similarity 0.81)
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.