🏢 School of Artificial Intelligence, Jilin University

Current LLM evaluation benchmarks often overlook the structural dependencies in multi-turn dialogues, treating them as simple concatenations of single-turn interactions. This approach neglects user in…