Research map / Large language models / Retrieval and knowledge
VSPO: Validating Semantic Pitfalls in Ontology via LLM-Based CQ Generation
Hyojun Choi, Seokju Hwang, Kyong-Ho Lee
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 |
|---|---|---|---|
| fine-tune LLaMA-3.1-8B-Instruct on misalignment-injected CQ data (VSPO) | LLaMA-3.1-8B-Instruct · CQsp (Types 1-3) | F1 (%): 24.3 -> 68.0 | Improved |
| VSPO vs GPT-4.1 on misalignment CQs | GPT-4.1 · CQsp (Types 1-3) | F1 (%): 40.1 -> 68.0 | Improved |
| VSPO vs GPT-4.1 on normal CQs | GPT-4.1 · CQnormal | F1 (%): 40.7 -> 52.1 | Improved |
| VSPO vs GPT-4.1 when axiom and definition are aligned | GPT-4.1 · CQnormal Type 4 | F1 (%): 65.6 -> 80.4 | Improved |
| VSPO vs GPT-4.1 on undefined axioms | GPT-4.1 · CQsp Type 2 | F1 (%): 36.1 -> 75.9 | Improved |
| evaluate on unseen ontology Pizza (leave-one-out) | VSPO · CQsp Pizza unseen | F1 (%): 68.0 -> 53.0 | Hurt |
| train more than 3 epochs (ablation) | VSPO LoRA fine-tuning · training | overfitting | Hurt |
Closest papers by meaning
- End-to-End Ontology Learning with Large Language Models (NeurIPS 2024 · similarity 0.85)
- ORACLE: Optimizing Reasoning Abilities of Large Language Models via Constraint-Led Synthetic Data Elicitation (AAAI 2026 · similarity 0.85)
- Diagnosing and Addressing Pitfalls in KG-RAG Datasets: Toward More Reliable Benchmarking (NeurIPS 2025 · similarity 0.84)
- DiscoveryBench: Towards Data-Driven Discovery with Large Language Models (ICLR 2025 · similarity 0.84)
- ThinkBench: Dynamic Out-of-Distribution Evaluation for Robust LLM Reasoning (NeurIPS 2025 · similarity 0.84)
- AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions (NeurIPS 2025 · 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.
