Add OSQB model

[ganeshpatelQB]

@liamdugan We want to evaluate our model, thanks!!!

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

OSQB

Release date: 2025-12-02

I've committed detailed results of this detector's performance on the test set to this PR.

On the RAID dataset as a whole (aggregated across all generation models, domains, decoding strategies, repetition penalties, and adversarial attacks), it achieved an AUROC of 99.60 and a TPR of 98.59% at FPR=5% and 95.19% at FPR=1%.
Without adversarial attacks, it achieved AUROC of 99.91 and a TPR of 99.71% at FPR=5% and 99.07% at FPR=1%.

If all looks well, a maintainer will come by soon to merge this PR and your entry/entries will appear on the leaderboard. If you need to make any changes, feel free to push new commits to this PR. Thanks for submitting to RAID!

[liamdugan]

Nice job on the submission! Let me know if you'd like us to merge your results

[liamdugan]

@ganeshpatelQB any updates? Happy to merge if you'd like or keep the PR open if you'd like to improve further

[ganeshpatelQB]

Hey @liamdugan , added a new model for the ealuations.

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

QuillBot

Release date: 2025-12-23

I've committed detailed results of this detector's performance on the test set to this PR.

On the RAID dataset as a whole (aggregated across all generation models, domains, decoding strategies, repetition penalties, and adversarial attacks), it achieved an AUROC of 99.24 and a TPR of 96.26% at FPR=5% and 88.26% at FPR=1%.
Without adversarial attacks, it achieved AUROC of 99.68 and a TPR of 98.88% at FPR=5% and 94.79% at FPR=1%.

If all looks well, a maintainer will come by soon to merge this PR and your entry/entries will appear on the leaderboard. If you need to make any changes, feel free to push new commits to this PR. Thanks for submitting to RAID!

[liamdugan]

@ganeshpatelQB let me know if you'd like to merge this one

[ganeshpatelQB]

Hey, thanks, @liamdugan, not yet.
Will give it another try.

[ganeshpatelQB]

Hey @liamdugan , added new prediction file for the evaluations.

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

QuillBot

Release date: 2025-01-27

I've committed detailed results of this detector's performance on the test set to this PR.

On the RAID dataset as a whole (aggregated across all generation models, domains, decoding strategies, repetition penalties, and adversarial attacks), it achieved an AUROC of 99.49 and a TPR of 98.97% at FPR=5% and 96.60% at FPR=1%.
Without adversarial attacks, it achieved AUROC of 99.45 and a TPR of 98.90% at FPR=5% and 97.04% at FPR=1%.

If all looks well, a maintainer will come by soon to merge this PR and your entry/entries will appear on the leaderboard. If you need to make any changes, feel free to push new commits to this PR. Thanks for submitting to RAID!

[ganeshpatelQB]

Hey @liamdugan , added new prediction file for the evaluations.

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

QuillBot

Release date: 2025-01-30

I've committed detailed results of this detector's performance on the test set to this PR.

On the RAID dataset as a whole (aggregated across all generation models, domains, decoding strategies, repetition penalties, and adversarial attacks), it achieved an AUROC of 99.62 and a TPR of 99.62% at FPR=5% and 98.10% at FPR=1%.
Without adversarial attacks, it achieved AUROC of 99.78 and a TPR of 99.72% at FPR=5% and 99.24% at FPR=1%.

If all looks well, a maintainer will come by soon to merge this PR and your entry/entries will appear on the leaderboard. If you need to make any changes, feel free to push new commits to this PR. Thanks for submitting to RAID!

[ganeshpatelQB]

Hey @liamdugan, can you merge the results?
I am assuming this will only publish the latest model and not the earlier ones!

[liamdugan]

Yep! Happy to do so. Congrats on the strong performance. (and yes it'll only publish the latest model)