AIDG: Evaluating Asymmetry Between Information Extraction and Containment in Multi-Turn Dialogue

arXiv:2602.17443v1 Announce Type: new Abstract: Evaluating the strategic reasoning capabilities of Large Language Models (LLMs) requires moving beyond static benchmarks to dynamic, multi-turn interactions. We …