UVA-IRLab Team Presents Multi-Turn RAG System for SemEval-2026 Conversational Question Answering Task
Researchers from UVA's Information Retrieval Lab submitted a system for SemEval-2026 Task 8 that combines learned sparse retrieval with LLM-based reranking to answer multi-turn conversational questions across four domains. The system is designed to handle unanswerable queries and maintain conversational context throughout the retrieval and generation pipeline. The work demonstrates how sparse retrieval methods can generalize effectively across diverse domains while leveraging LLM capabilities for complex reasoning tasks.
The UVA-IRLab team developed a multi-turn retrieval-augmented generation (RAG) pipeline for SemEval-2026 Task 8, which evaluates conversational question-answering systems across finance, cloud documentation, government, and Wikipedia domains. Their approach prioritizes learned sparse retrieval as the primary retrieval method, supplemented by LLM-based pointwise and listwise reranking. The system uses long-context LLM capabilities for conversational query rewriting and final response generation, with each component conditioned on the full conversational history. This multi-step design aims to improve robustness across diverse domains while handling cases where the available collection contains insufficient evidence for complete answers. The paper presents results across five figures and six tables documenting the system's performance.
What different sources said
- arXiv cs.CLCenter
uva-irlab-conv at SemEval-2026 Task 8: Multi-Turn RAG with Learned Sparse Retrieval and Listwise Reranking
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.