Does Chain of Thought cause models to converge more?

2

LLMs and many generative models appear to be converging in representations (https://arxiv.org/abs/2405.07987 https://arxiv.org/abs/2504.08775) and persona (https://cichicago.substack.com/p/persona-collapse https://arxiv.org/abs/2510.22954).

Does Chain of Thought cause more or less convergence?

llms persona convergence Implemented:https://github.com/Hypogenic-AI/chain-of-thought-conv-b564-claude Implemented:https://github.com/Hypogenic-AI/chain-of-thought-conv-1189-codex

Chat

If you are inspired by this idea, you can reach out to the authors for collaboration or cite it:

@misc{holtzman-does-chain-of-2026,
  author = {Holtzman, Ari},
  title = {Does Chain of Thought cause models to converge more?},
  year = {2026},
  url = {https://hypogenic.ai/ideahub/idea/vbUl35tUGNzskaqwPcCq}
}

Comments (0)

Please sign in to comment on this idea.

No comments yet. Be the first to share your thoughts!