Research map / Large language models / Retrieval and knowledge
ALEX:A Light Editing-knowledge Extractor
Minghu Wang, ShuLiang Zhao, Yuanyuan Zhao, Hongxia Xu
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 |
|---|---|---|---|
| add ALEX hierarchical retrieval to PokeMQA | PokeMQA · MQuAKE-CF-3K (DeepSeek) | MultiHop / HopWise accuracy: 41.90 -> 51.33 MA; 29.83 -> 43.60 HA | Improved |
| add ALEX to MeLLo | MeLLo · MQuAKE-T (Llama 2 7B) | MultiHop accuracy: 59.26 -> 68.2 | Improved |
| add ALEX to PokeMQA | PokeMQA · MQuAKE-Hard (DeepSeek) | HopWise accuracy: 25.78 -> 74.35 | Improved |
| add ALEX to PokeMQA on a small backbone | PokeMQA · MQuAKE-CF-3K (Llama 2 7B) | MultiHop accuracy: HA 22.91 -> 36.80; MA 34.20 -> 33.43 | Mixed |
| add ALEX to PokeMQA with Llama3.1 8B | PokeMQA · MQuAKE-CF-3K-v2 (Llama3.1 8B) | MultiHop accuracy: 41.00 -> 39.32 | Hurt |
| add DEA and IQS together versus neither (ablation) | ALEX · MQuAKE-T (DeepSeek) | MultiHop accuracy: 70.53 -> 87.33 | Improved |
| add IQS on top of DEA (ablation) | ALEX · MQuAKE-CF-3K-v2 (DeepSeek) | MultiHop accuracy: 48.17 -> 53.50 | Improved |
| add DEA on top of IQS (ablation) | ALEX · MQuAKE-Hard (DeepSeek) | MultiHop accuracy: 74.84 -> 79.20 | Improved |
| add DEA alone versus IQS alone (ablation) | ALEX · MQuAKE-CF-3K-v2 vs MQuAKE-Hard (DeepSeek) | MultiHop accuracy: v2: DEA 48.17 vs IQS 41.75; Hard: DEA 67.55 vs IQS 74.84 | Mixed |
| cluster edits into 12 clusters before retrieval (ablation) | ALEX · MQuAKE-CF-3K-v2 | average retrieval count: 2764 -> 368 | Improved |
| vary number of clusters K from 5 to 20 (ablation) | ALEX · MQuAKE datasets | Cluster ACC / Retrieval ACC | No clear effect |
| add ALEX to MeLLo with Llama3.1 8B | MeLLo · MQuAKE-CF-3K (Llama3.1 8B) | MultiHop accuracy: 8.36 -> 9.61 | Improved |
Closest papers by meaning
- Knowledge Editing with Dynamic Knowledge Graphs for Multi-Hop Question Answering (AAAI 2025 · similarity 0.90)
- Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing (ICML 2025 · similarity 0.88)
- Autofocus Retrieval: An Effective Pipeline for Multi-Hop Question Answering With Semi-Structured Knowledge (NeurIPS 2026 · similarity 0.87)
- Harnessing Large Language Models for Knowledge Graph Question Answering via Adaptive Multi-Aspect Retrieval-Augmentation (AAAI 2025 · similarity 0.87)
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 2025 · similarity 0.87)
- AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models (ICLR 2025 · similarity 0.86)
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.
