Tarski attack shows LLM probes cannot detect truth

A blog post applies Tarski's undefinability theorem to LLM probing, arguing that linear probes cannot reliably detect truth in model representations. The critique suggests fundamental limits to interpretability via probes.