One of the reasons LLMs write annoyingly is that they don't seem to think of the audience in a vague way. I bet have the model manually simulate the audience in its CoT every sentence or so would help a lot.
If you are inspired by this idea, you can reach out to the authors for collaboration or cite it:
@misc{holtzman-llms-dont-do-2026,
author = {Holtzman, Ari},
title = {LLMs Don't Do Much ToM in Writing},
year = {2026},
url = {https://hypogenic.ai/ideahub/idea/LpTXdOL1sSqjyqLWxGCU}
}Please sign in to comment on this idea.
No comments yet. Be the first to share your thoughts!