Frontier LLMs Fail at Counterfactual Reasoning: Best Score Just 64.6%

AryHHAry · x · 2026-09-01

A new paper from Zhejiang University, Alibaba, and City University of Hong Kong reveals that while LLMs can give convincing answers to "What If" counterfactual questions, the underlying causal chains are often incoherent.

Research Core:

Results:

Six frontier models were tested, with none achieving saturation:

Key Failure Modes:

Original post →

More from Research

Research channel →