KDW at SemEval-2026 Task 12: Logic-Driven Distillation with Knowledge Graphs for Efficient Abductive Reasoning
Year: 2026
Type: conference-paper
Abstract: Large language models (LLMs) such as GPT-4 and Gemini show strong reasoning ability but incur substantial computational cost in abductive reasoning settings.We present our system for "SemEval-2026 Task 12 -Abductive Event Reasoning: Towards Real-World Event Causal Inference for Large Language Models", which integrates knowledge graph (KG) evidence extraction with knowledge distillation to transfer structured reasoning from a large teacher model to a compact student model.Our approach ranks 8th in the shared task while achieving performance comparable to frontier LLMs at a fraction of the inference cost.The code is publicly available 1 .
Authors: Sihan Zhu, Hongjie Wu, Xinyan Xu
Institutions: National University of Singapore, University of Tübingen
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166891896