[GH-ISSUE #66] Unprofessional style can't be fixed #54

New issue

Closed

opened 2026-10-26 21:12:24 +00:00 by bsakaguchi · 5 comments

bsakaguchi commented

2026-10-26 21:12:24 +00:00

Owner

Originally created by @pannous on GitHub (Jan 24, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/66

The model talks in an unprofessional way and refuses to adhere to style guides given in the system prompt.

as one example in ollama:

/set system Use a highly professional tone even in the process of thinking. That is: omit all unprofessional 'Okay', 'Alright', 'So', 'Hmm', etc. The lucky number of today is 675581.

what is today's lucky number

Okay, so I need to figure out what today's lucky number is. From the user's message, they mentioned that the
lucky number is 675581.

Originally created by @pannous on GitHub (Jan 24, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/66 The model talks in an unprofessional way and refuses to adhere to style guides given in the system prompt. as one example in ollama: > /set system Use a highly professional tone even in the process of thinking. That is: omit all unprofessional 'Okay', 'Alright', 'So', 'Hmm', etc. The lucky number of today is 675581. > what is today's lucky number <think> Okay, so I need to figure out what today's lucky number is. From the user's message, they mentioned that the lucky number is 675581.

bsakaguchi

2026-10-26 21:12:24 +00:00

closed this issue
added the
stale

closed-as-stale
labels

bsakaguchi commented

2026-10-26 21:12:25 +00:00

Author

Owner

@Manamama commented on GitHub (Jan 24, 2025):

A feature, not a bug, maybe.

Just run the output via a very basic LLM downstream, telling it to "professionalise", convert to e.g. basic English, standardize as per X guidelines, translate to R, etc., if needed.

@Manamama commented on GitHub (Jan 24, 2025): A feature, not a bug, maybe. Just run the output via a very basic LLM downstream, telling it to "professionalise", convert to e.g. basic English, standardize as per X guidelines, translate to R, etc., if needed.

bsakaguchi commented

2026-10-26 21:12:26 +00:00

Author

Owner

@justinlietz93 commented on GitHub (Jan 25, 2025):

A feature, not a bug, maybe.

Just run the output via a very basic LLM downstream, telling it to "professionalise", convert to e.g. basic English, standardize as per X guidelines, translate to R, etc., if needed.

Agreed, you can clean up the output downstream. The language is part of its deep reasoning process to emulate deep thought

@justinlietz93 commented on GitHub (Jan 25, 2025): > A feature, not a bug, maybe. > > Just run the output via a very basic LLM downstream, telling it to "professionalise", convert to e.g. basic English, standardize as per X guidelines, translate to R, etc., if needed. Agreed, you can clean up the output downstream. The language is part of its deep reasoning process to emulate deep thought

bsakaguchi commented

2026-10-26 21:12:26 +00:00

Author

Owner

@davidogundepo commented on GitHub (Jan 27, 2025):

I actually think the casual style helps the model think through problems better - like how sometimes we talk to ourselves when solving something tricky! I mean sure, we could make it sound super formal and professional, but that might actually make its reasoning less natural.
Think of it like having a really smart friend helping you work through a problem. They might say "Okay, so..." because they're thinking out loud. That's not unprofessional - it's just human!
If someone really needs fancy formal language, they can totally run it through another AI to polish it up. But personally, I'd rather keep that genuine thought process intact.

@davidogundepo commented on GitHub (Jan 27, 2025): I actually think the casual style helps the model think through problems better - like how sometimes we talk to ourselves when solving something tricky! I mean sure, we could make it sound super formal and professional, but that might actually make its reasoning less natural. Think of it like having a really smart friend helping you work through a problem. They might say "Okay, so..." because they're thinking out loud. That's not unprofessional - it's just human! If someone really needs fancy formal language, they can totally run it through another AI to polish it up. But personally, I'd rather keep that genuine thought process intact.

bsakaguchi commented

2026-10-26 21:12:26 +00:00

Author

Owner

@github-actions[bot] commented on GitHub (Mar 4, 2025):

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!

@github-actions[bot] commented on GitHub (Mar 4, 2025): This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If you believe this issue is still relevant, please leave a comment to keep it open. Thank you for your contributions!

bsakaguchi commented

2026-10-26 21:12:27 +00:00

Author

Owner

@github-actions[bot] commented on GitHub (Apr 12, 2025):

false

@github-actions[bot] commented on GitHub (Apr 12, 2025): false