Research map / Large language models / Retrieval and knowledge
Improving Natural Language Inference Using External Knowledge in the Science Questions Domain
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 a ConceptNet Concepts Only graph model with CN-PPMI embeddings to match-LSTM (GRU) | match-LSTM (GRU) · SciTail test | accuracy: 84.2 -> 85.2 | Helped |
| Replace LSTMs with GRUs in match-LSTM | match-LSTM · SciTail test | accuracy: 84.1 -> 84.2 | Helped |
| Add WordNet features to match-LSTM | match-LSTM (GRU) · SciTail test | accuracy: 84.2 -> 84.3 | Helped |
| Use ConceptNet rather than WordNet or DBpedia as the graph knowledge source (ablation) | match-LSTM+GconAttn · SciTail dev | accuracy: WordNet 87.6; DBpedia 87.3; ConceptNet 88.6 | Helped |
| Use graph-only prediction rather than combining graph and text (ablation) | match-LSTM+GconAttn with Concepts Only graph · SciTail dev | accuracy: 72.3 vs 87.2 | Hurt |
| Add One-Hop neighbors to Concepts Only graphs (ablation) | Graph-only GconAttn · SciTail dev | accuracy: 72.3 -> 68.2 | Hurt |
| Add One-Hop neighbors to Concepts Only graphs (ablation) | Text+Graph GconAttn · SciTail dev | accuracy: 87.2 -> 87.3 | No clear effect |
| Add Two-Hop paths to Concepts Only graphs (ablation) | Text+Graph GconAttn · SciTail dev | accuracy: 87.2 -> 87.3 | No clear effect |
| Replace CN-PPMI with TransH embeddings (ablation) | match-LSTM+Gmatch-LSTM · SciTail dev | accuracy: 89.6 -> 87.3 | Hurt |
| Replace CN-PPMI with CompleX embeddings (ablation) | match-LSTM+Gmatch-LSTM · SciTail dev | accuracy: 89.6 -> 88.3 | Hurt |
| Replace Gmatch-LSTM with GconAttn using CN-PPMI embeddings (ablation) | match-LSTM graph-text model · SciTail dev | accuracy: 89.6 -> 88.6 | Hurt |
| Add graph predictions as a complementary source and choose the correct answer between text and graph models (ablation) | Text-only match-LSTM plus graph model · SciTail dev | accuracy: 88.5 versus at least 91.6 | Helped |
Closest papers by meaning
- Enhancing Natural Language Inference Using New and Expanded Training Data Sets and New Learning Models (AAAI 2020 · similarity 0.89)
- Natural Language Inference in Context – Investigating Contextual Reasoning over Long Texts (AAAI 2021 · similarity 0.89)
- Probing Natural Language Inference Models through Semantic Fragments (AAAI 2020 · similarity 0.89)
- Debate on Graph: A Flexible and Reliable Reasoning Framework for Large Language Models (AAAI 2025 · similarity 0.87)
- Harnessing Large Language Models for Knowledge Graph Question Answering via Adaptive Multi-Aspect Retrieval-Augmentation (AAAI 2025 · similarity 0.86)
- Linguini: A benchmark for language-agnostic linguistic reasoning (NeurIPS 2025 · similarity 0.85)
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.