Autologic Fun with LLMs

I asked four frontier LLMs – two from OpenAI and two from Anthropic – to write a paragraph such that the paragraph is densely packed with autologic words and phrases and the paragraph itself is autologic. The generated paragraphs were judged by a human and by an LLM judge. The two Anthropic models took first and second place; the two OpenAI models came in last.

Click here for the full report (PDF).