WILT: A Multi-Turn, Memorization-Robust Inductive Logic Benchmark for LLMs

Eryk Banatt, Jonathan Cheng, Skanda Vaidyanath +1

While large language models have shown impressive capabilities across a wide range of domains, they still encounter significant challenges in reasoning tasks that require gathering…