Team DU achieved first place in the COLIEE 2026 statute entailment task by employing a cross-architecture ensemble of nine models. Their approach also yielded strong results in other legal information processing tasks, including tort prediction and legal case entailment, by utilizing multi-view systems and prompt modifications. The team found that combining different models and retrieval-augmented prompting techniques proved most effective across various legal information processing challenges. AI
IMPACT Demonstrates effectiveness of cross-architecture LLM ensembling and advanced prompting for specialized legal tasks.
RANK_REASON Academic paper detailing participation and results in a legal information processing competition.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
