Date: 2026-10-31

Team “tus-nlp,” including Akira Ise (a fourth-year undergraduate student in our lab), placed third in the LitTraceQA competition held at GroundLM 2026, a workshop co-located with EMNLP 2026.

LitTraceQA challenges AI systems to retrieve scientific papers, locate supporting evidence, and answer questions accurately.

Our team developed Lit3R, a system that retrieves papers relevant to a question and expands its search using relationships between papers. It then combines evidence from multiple papers to generate an answer.

Competition report:
Findings of the First GroundLM Shared Tasks: Evaluating Grounded Language Models Across Visual and Scientific Evidence

System paper:
Lit3R: Retrieve–Relate–Read for Evidence-Grounded Question Answering over Scientific Literature

Updated: