TALES: Text Adventure Learning Environment Suite
Date:
Oral presentation of TALES: Text Adventure Learning Environment Suite — a progressive stress test of LLMs’ baseline agentic capabilities, spanning 122 tasks across which 49 models were benchmarked.
Joint work with Xingdi “Eric” Yuan, Ziang Xiao, Prithviraj Ammanabrolu, and Marc-Alexandre Côté.
Project page: microsoft.github.io/tale-suite