AI RESEARCH
Investigating Advanced Reasoning of Large Language Models via Black-Box Environment Interaction
arXiv CS.AI
•
ArXi:2508.19035v2 Announce Type: replace Existing tasks fall short in evaluating reasoning ability of Large Language Models (LLMs) in an interactive, unknown environment. This deficiency leads to the isolated assessment of deductive, inductive, and abductive reasoning, neglecting the integrated reasoning process that is indispensable for human-like discovery learning. We