Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces!

Intermediate token generation (ITG), where a model produces output before the solution, has been proposed as a method to improve the performance of language models on reasoning tasks. These intermediate tokens have been called "reasoning traces" or even "thoughts" -- implicitly anthropomorphizing the model, implying these tokens resemble steps a human might take when solving a challengingthis http URLthis paper, we present evidence that this anthropomorphization isn't a harmless metaphor, and instead is quite dangerous -- it confuses the nature of these models and how to use them effectively, and leads to questionable research.

@article{kambhampati2025_2504.09762,
  title={ Stop Anthropomorphizing Intermediate Tokens as Reasoning/Thinking Traces! },
  author={ Subbarao Kambhampati and Kaya Stechly and Karthik Valmeekam and Lucas Saldyt and Siddhant Bhambri and Vardhan Palod and Atharva Gundawar and Soumya Rani Samineni and Durgesh Kalwar and Upasana Biswas },
  journal={arXiv preprint arXiv:2504.09762},
  year={ 2025 }
}