Research map / Large language models / Retrieval and knowledge
PKR-QA: A Benchmark for Procedural Knowledge Reasoning with Knowledge Module Learning
Thanh-Son Nguyen, Hong Yang, Tzeh Yuan Neoh, Hao Zhang, Ee Yeo Keat, Basura Fernando
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 P.VRL predicted step/task categories to VLM input (ablation) | VLMs · PKR-QA | accuracy: Qwen2.5-VL 59.6 -> 69.4 | Improved |
| fine-tune VLMs on KG triplets (ablation) | VLMs · PKR-QA | accuracy: Qwen2.5-VL 59.6 -> 67.3 | Improved |
| fine-tune VLMs on 1,700 QA pairs (ablation) | VLMs · PKR-QA | accuracy: Qwen2.5-VL 59.6 -> 73.6 | Improved |
| add KG training on top of QA training (ablation) | VLMs · PKR-QA | accuracy: Qwen2.5-VL 73.6 -> 74.2 | Improved |
| use KML-CLIP with KG+QA training instead of VLM or KG-embedding baselines | neurosymbolic methods · PKR-QA | accuracy: best VLM 74.2 -> 78.1 | Improved |
| add KG+QA training to KML-CLIP (versus KG-training only) (ablation) | KML-CLIP · PKR-QA | accuracy: 75.3 -> 78.1 | Improved |
| learn KML embeddings from scratch instead of frozen CLIP (ablation) | KML · PKR-QA | accuracy: KG: 74.6 -> 73.5; KG+QA: 76.7 -> 77.4 | Mixed |
| initialize KG embedding baselines with CLIP (ablation) | TransE/TransH/RotatE · PKR-QA | accuracy: TransE 63.6 -> 56.8; RotatE 41.6 -> 48.8 | Mixed |
| train KML modules on PKG relations rather than QA only (ablation) | KML-F-CLIP · PKR-QA | mean accuracy: 59.3 -> 71.6 (accuracy KG-training) | Improved |
| generate multiple programs instead of one (ablation) | KML · PKR-QA | accuracy | No clear effect |
| train modules on generic WikiHow KG instead of domain PKG (ablation) | KML-F-CLIP · PKR-QA | mean accuracy: 74.9 (WikiHow KG+QA) vs PKG-based 75.3 | Hurt |
Closest papers by meaning
- KBQA-R1: Reinforcing Large Language Models for Knowledge Base Question Answering (ICML 2026 · similarity 0.87)
- Procedural Knowledge in Pretraining Drives Reasoning in Large Language Models (ICLR 2025 · similarity 0.87)
- ProgRAG: Hallucination-Resistant Progressive Retrieval and Reasoning over Knowledge Graphs (AAAI 2026 · similarity 0.86)
- K-DeCore: Facilitating Knowledge Transfer in Continual Structured Knowledge Reasoning via Knowledge Decoupling (NeurIPS 2025 · similarity 0.86)
- Improving Complex Reasoning over Knowledge Graph with Logic-Aware Curriculum Tuning (AAAI 2025 · similarity 0.85)
- Debate on Graph: A Flexible and Reliable Reasoning Framework for Large Language Models (AAAI 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.
