Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!

arxiv.org

saved by: 0 people
related read on: OpenAI’s new reasoning technique alarms AI safety experts from the discussion