Research map / Large language models / Retrieval and knowledge
LEGO: Latent Execution-Guided Reasoning for Multi-Hop Question Answering on Knowledge Graphs
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 |
|---|---|---|---|
| add stochastic hard EM (min-pooling over candidate query losses) to training (ablation) | LEGO query synthesizer · CWQ and WQSP | Hits@1: up to 1.6 points | Helped |
| add execution-guided search space pruning (branch pruner and top-50 relation pruner) during candidate query search (ablation) | LEGO training candidate search · WQSP (50% KG) | Hits@1: 48.5 vs 13.2 | Helped |
| LEGO latent execution without natural-language input versus KG traversal with ground truth query | LEGO on MetaQA · MetaQA 50% KG | Hits@1: 63.8 vs 51.5 (overall) | Helped |
| use LEGO relative to EmbedKGQA on single-hop questions | LEGO on MetaQA 1-hop · MetaQA 50% KG 1-hop | Hits@1: 69.3 vs 70.6 | Hurt |
| LEGO versus strongest baseline on multi-hop questions | LEGO on MetaQA 2-hop and 3-hop · MetaQA 50% KG | Hits@1: 57.8 vs 54.3 (2-hop); 63.8 vs 59.7 (3-hop) | Helped |
| LEGO on WQSP with 70% missing KG edges | LEGO · WQSP (70% missing) | Hits@1: 38.0 vs 34.6 (Pullnet) and 31.4 (EmbedKGQA) | Helped |
Closest papers by meaning
- Knowledge Editing with Dynamic Knowledge Graphs for Multi-Hop Question Answering (AAAI 2025 · similarity 0.90)
- Tree-of-Reasoning Question Decomposition for Complex Question Answering with Large Language Models (AAAI 2024 · similarity 0.89)
- Debate on Graph: A Flexible and Reliable Reasoning Framework for Large Language Models (AAAI 2025 · similarity 0.89)
- Diagnosing and Addressing Pitfalls in KG-RAG Datasets: Toward More Reliable Benchmarking (NeurIPS 2025 · similarity 0.89)
- KG-Guard: Graph-Based Hallucination Detection for Knowledge Base Question Answering (NeurIPS 2026 · similarity 0.89)
- Question Calibration and Multi-Hop Modeling for Temporal Question Answering (AAAI 2024 · similarity 0.88)
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.