Saved 17 images to 'pdf_pages'

=== Sent 2 image(s): pages [1, 2] ===
Rank1 was trained on **635,264 examples** of R1 reasoning traces. This number is given towards the end of the "Data Preparation" section on page 2:

> "After all generation was done, our dataset has 635,264 examples of R1 generations..."

So, 635,264 reasoning traces from R1 were sourced to train Rank1.
This response took 1686 tokens

=== Sent 4 image(s): pages [1, 2, 3, 4] ===
The number of reasoning traces from R1 that were sourced to train Rank1 is **635,264**.

This is found on page 2 in the "Data Preparation" section:
> "After all generation was done, our dataset has 635,264 examples of R1 generations..."
This response took 3199 tokens

=== Sent 6 image(s): pages [1, 2, 3, 4, 5, 6] ===
The number of reasoning traces from R1 that were sourced to train Rank1 is **635,264**.

You can find this information on page 2, section "2.1 Data Preparation", where it states:
> "After all generation was done, our dataset has 635,264 examples of R1 generations, where R1 labeled 62.9% as relevant and 37.1% as non-relevant."
This response took 4758 tokens

=== Sent 8 image(s): pages [1, 2, 3, 4, 5, 6, 7, 8] ===
The number of reasoning traces from R1 that were sourced to train Rank1 is **635,264 examples**. 

This can be found in Section 2.1 ("Data Preparation") on page 2, where it says:

> "After all generation was done, our dataset has 635,264 examples of R1 generations, where R1 labeled 62.9% as relevant and 37.1% as non-relevant."

So, **635,264 R1 reasoning traces** were sourced for training Rank1.
This response took 6309 tokens

=== Sent 10 image(s): pages [1, 2, 3, 4, 5, 6, 7, 8, 9, 10] ===
The number of reasoning traces from R1 that were sourced to train Rank1 is **635,264**.

You can find this on page 2, where it is stated: 
> "After all generation was done, our dataset has 635,264 examples of R1 generations..."

This corresponds to the number of R1-generated reasoning chains used for Rank1’s training.
This response took 7807 tokens