Research map / Multi-agent systems and games / Human-AI and goal-conditioned
A Multi-Agent Communication Framework for Question-Worthy Phrase Extraction and Question Generation
Siyuan Wang, Zhongyu Wei, Zhihao Fan, Yang Liu, Xuanjing Huang
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 |
|---|---|---|---|
| pass question information from the generation agent to the extraction agent (E_MP vs E_Local) (ablation) | phrase extractor · SQuAD sentence-level test phrase extraction | exact match: 24.27% -> 35.77% | Improved |
| pass question information from the generation agent (E_MP vs E_Local) (ablation) | phrase extractor · SQuAD sentence-level test phrase extraction | F1: 38.63 -> 46.71 | Improved |
| pointer-network extractor instead of Stanford NER | phrase extractor · SQuAD sentence-level test phrase extraction | F1: 17.33 -> 38.63 | Improved |
| message passing reduces overlapping phrases (ablation) | phrase extractor · SQuAD sentence-level test phrase extraction | sentences with overlapped phrases: 32.37% -> 22.05% | Improved |
| message passing discards mis-extracted phrases (ablation) | phrase extractor · SQuAD sentence-level test phrase extraction | avg. number of extracted phrases: 1.43 -> 1.38 | Mixed |
| use extracted phrases from E_MP trained jointly instead of E_Local phrases (ablation) | question generator · SQuAD sentence-level test question generation | BLEU-4: 9.76 -> 10.56 | Improved |
| use extracted phrases from E_MP instead of E_Local phrases (ablation) | question generator · SQuAD sentence-level test question generation | BLEU-1: 44.36 -> 45.70 | Improved |
| use extracted phrases from E_MP instead of E_Local phrases (ablation) | question generator · SQuAD sentence-level test question generation | ROUGE-L: 37.00 -> 38.09 | Improved |
| feed NER-extracted phrases as extra input vs no phrase information (ablation) | question generator · SQuAD sentence-level test question generation | BLEU-4: 9.05 -> 9.22 | No clear effect |
| feed E_MP phrases instead of no phrase information (ablation) | question generator · SQuAD sentence-level test question generation | BLEU-4: 9.05 -> 10.56 | Improved |
| feed E_MP phrases instead of no phrase information (ablation) | question generator · SQuAD sentence-level test question generation | METEOR: 14.63 -> 15.76 | Improved |
| use ground-truth answers instead of extracted phrases (upper bound) (ablation) | question generator · SQuAD sentence-level test question generation | BLEU-4: 10.56 -> 11.81 | Improved |
Closest papers by meaning
- Chain of Agents: Large Language Models Collaborating on Long-Context Tasks (NeurIPS 2024 · similarity 0.86)
- AgentHop: A Diagnostic Benchmark for Agentic Multi-Hop Scientific Question Answering (NeurIPS 2026 · similarity 0.84)
- Knowledge Tagging with Large Language Model Based Multi-Agent System (AAAI 2025 · similarity 0.84)
- Divide-and-Conquer: Tree-structured Strategy with Answer Distribution Estimator for Goal-Oriented Visual Dialogue (AAAI 2025 · similarity 0.84)
- How Teachers Can Use Large Language Models and Bloom’s Taxonomy to Create Educational Quizzes (AAAI 2024 · similarity 0.83)
- MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains (ICLR 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.
