[GH-ISSUE #205] Clarification on "Cold-Start Data" #145

Closed
opened 2026-10-26 21:15:20 +00:00 by bsakaguchi · 4 comments
Owner

Originally created by @jigarvyasidea on GitHub (Jan 31, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/205

The README mentions "cold-start data" before RL, but doesn't elaborate on what this data is or how it is used. An issue requesting a description of this data would be beneficial.

Originally created by @jigarvyasidea on GitHub (Jan 31, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/205 The README mentions "cold-start data" before RL, but doesn't elaborate on what this data is or how it is used. An issue requesting a description of this data would be beneficial.
bsakaguchi 2026-10-26 21:15:20 +00:00
Author
Owner

@wuwowuyi commented on GitHub (Feb 3, 2025):

I'd also like to know more details on how the data is collected.
For example, the paper mentions "using few-shot prompting with a long CoT as an example", which model is prompted? Is it R1-zero? or DeepSeek-v3?

<!-- gh-comment-id:2629819758 --> @wuwowuyi commented on GitHub (Feb 3, 2025): I'd also like to know more details on how the data is collected. For example, the paper mentions "using few-shot prompting with a long CoT as an example", which model is prompted? Is it R1-zero? or DeepSeek-v3?
Author
Owner

@michael-tj commented on GitHub (Feb 12, 2025):

I'd also like to know more details on how the data is collected. For example, the paper mentions "using few-shot prompting with a long CoT as an example", which model is prompted? Is it R1-zero? or DeepSeek-v3?

According to the paper sentences. long CoT data maybe from R1-zero.
paper:

Image

<!-- gh-comment-id:2652908771 --> @michael-tj commented on GitHub (Feb 12, 2025): > I'd also like to know more details on how the data is collected. For example, the paper mentions "using few-shot prompting with a long CoT as an example", which model is prompted? Is it R1-zero? or DeepSeek-v3? According to the paper sentences. long CoT data maybe from R1-zero. paper: ![Image](https://github.com/user-attachments/assets/061299a1-adb4-421a-8dde-499f5a737dc9)
Author
Owner

@github-actions[bot] commented on GitHub (Apr 5, 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!

<!-- gh-comment-id:2779931452 --> @github-actions[bot] commented on GitHub (Apr 5, 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!
Author
Owner

@github-actions[bot] commented on GitHub (May 8, 2025):

false

<!-- gh-comment-id:2860886145 --> @github-actions[bot] commented on GitHub (May 8, 2025): false
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepSeek-R1#145
No description provided.