Expected output
#3
by
sanderland
- opened
Can you confirm the expected output is as below, it seems it may be a copy-paste error from the non-40M version
# Expected output:
# Score for response 1: 23.125
# Score for response 2: 3.578125
Yes, it's a copy-paste from the non-40M version since we use the same README for all models. For this specific model, the output we receive is:
Score for response 1: 38.25
Score for response 2: 11.5625
Thank you! Would you also have these numbers for Skywork/Skywork-Reward-V2-Qwen3-8B
?
Also do you plan to release the data for the new series of models?
For Skywork-Reward-V2-Qwen3-8B, the scores are:
Score for response 1: 11.75
Score for response 2: -0.9609375
Regarding the release of the data, yes, we are awaiting internal approval and working through licensing issues. We'd love to have them released once these are resolved.
Amazing, thank you so much.
sanderland
changed discussion status to
closed