AILS-NTUA at SemEval-2026 Task 8: Evaluating Multi-Turn RAG Conversations
This work addresses multi-turn RAG evaluation for NLP researchers, but it is incremental as it builds on existing methods for a specific competition task.
The paper tackled the problem of multi-turn retrieval-augmented generation by proposing a unified system that ranked 1st in passage retrieval with a 20.5% improvement in nDCG@5 and 2nd in response generation.
We present the AILS-NTUA system for SemEval-2026 Task 8 (MTRAGEval), addressing all three subtasks of multi-turn retrieval-augmented generation: passage retrieval (A), reference-grounded response generation (B), and end-to-end RAG (C). Our unified architecture is built on two principles: (i) a query-diversity-over-retriever-diversity strategy, where five complementary LLM-based query reformulations are issued to a single corpus-aligned sparse retriever and fused via variance-aware nested Reciprocal Rank Fusion; and (ii) a multistage generation pipeline that decomposes grounded generation into evidence span extraction, dual-candidate drafting, and calibrated multi-judge selection. Our system ranks 1st in Task A (nDCG@5: 0.5776, +20.5% over the strongest baseline) and 2nd in Task B (HM: 0.7698). Empirical analysis shows that query diversity over a well-aligned retriever outperforms heterogeneous retriever ensembling, and that answerability calibration-rather than retrieval coverage-is the primary bottleneck in end-to-end performance.