Papers with structured intermediate representation
PosterForest: Hierarchical Multi-Agent Collaboration for Scientific Poster Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for scientific poster generation lack hierarchical document understanding and coherent content-layout planning. |
| Approach: | They propose a training-free framework for scientific poster generation that captures document hierarchy and semantics across multiple levels. |
| Outcome: | The proposed framework outperforms existing methods in both automatic and human evaluations without additional training or domain-specific supervision. |
Knowledge-Centric Templatic Views of Documents (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Prior work on document generation has tackled the creation of each separate format as a different task, leading to fragmented learning processes, redundancy in models and methods, and disjointed evaluation. |
| Approach: | They propose a method that unifies the generation and evaluation of templatic views of documents in multiple formats. |
| Outcome: | The proposed method improves performance for heterogeneous downstream applications while reducing the need for task specific evaluation metrics. |
SocraticKG: Knowledge Graph Construction via QA-Driven Fact Extraction (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to construct knowledge graphs struggle with factual coverage and information loss. |
| Approach: | They propose an automated KG construction method that introduces question-answer pairs as a structured intermediate representation to unfold document-level semantics prior to triple extraction. |
| Outcome: | The proposed method achieves superior factual retention while maintaining high structural cohesion even as extracted knowledge volume substantially expands. |