Open access
2026
CultRAG at SemEval-2026 Task 7: Hybrid Sparse-Dense Retrieval with Entity-Centric Knowledge Bases for Cultural MCQ Answering
The core finding is that RAG hurts rather than helps: the LLM-only baseline achieves 78.6% accuracy, outperforming the full system at 78.5% (McNemar’s test, p = 0 . 962).
Aditya Singh, Rickarya Das
· SemEval@ACL · 1 citation