Right for the Wrong Reasons: Epistemic Regret Minimization for Causal Rung Collapse in LLMs

Researchers identify a fundamental flaw in large language models called 'Rung Collapse' where AI systems achieve correct answers through flawed causal reasoning that fails under distribution shifts. They propose Epistemic Regret Minimization (ERM) as a solution that penalizes incorrect reasoning processes independently of task success, showing 53-59% recovery of reasoning errors in experiments across six frontier LLMs.

• →Large language models suffer from 'Rung Collapse' where they cannot distinguish between association and causation, leading to brittle reasoning.
• →Current autoregressive training reinforces correct answers obtained through incorrect causal models, creating 'Aleatoric Entrenchment'.
• →Epistemic Regret Minimization (ERM) addresses this by penalizing reasoning errors independently of whether the final answer is correct.
• →Even advanced reasoning-enhanced models like GPT-5.2 show persistent causal reasoning failures with only 3.7% success rates.
• →ERM feedback successfully recovered 53-59% of entrenched reasoning errors where traditional outcome-based feedback failed.