Add Plagiat AI Detector RAID submission

[cristeamarinescu-cpu]

Adds predictions and metadata for Plagiat AI Detector.

• predictions.json contains 56,000 {id, score} records.
• Predictions were generated on RAID-test no-adversarial using the staged detector.
• metadata.json includes contact: cristea.marinescu@ascendia.ro.
• No results.json is included; expecting the RAID bot to evaluate and generate it.

[cristeamarinescu-cpu]

Hello, could you please approve the workflow run for this RAID leaderboard submission when convenient? Thank you in advance.

[github-actions]

It looks like this eval run failed. Please check the workflow logs to see what went wrong, then push a new commit to your PR to rerun the eval.

[liamdugan]

Apologies @cristeamarinescu-cpu seems like there's an issue on our end with the evaluation bot (GitHub updated some default permission setting). Will follow up on this tonight.

[github-actions]

It looks like this eval run failed. Please check the workflow logs to see what went wrong, then push a new commit to your PR to rerun the eval.

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

Plagiat AI Detector

Release date: 2026-07-26

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

Warning

No aggregate score across all settings is reported here as some domains/generator models/decoding strategies/repetition penalties/adversarial attacks were not included in the submission. This submission will not appear in the main leaderboard; it will only be visible within the splits in which all samples were evaluated.

Warning

No aggregate score across all non-adversarial settings is reported here as some domains/generator models/decoding strategies/repetition penalties were not included in the submission.

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]

@cristeamarinescu-cpu bot is fixed now, seems like there were some predictions missing from your submission. Can you double check that you have outputs for every text in the dataset?

[cristeamarinescu-cpu]

Hello @liamdugan — you were right. Sorry for the long silence.

Our first submission only covered the 56,000 rows of test_none.csv; we had mistaken that file for the full test set. Since compute_scores(..., require_complete=True) also needs the adversarially-attacked human rows of each domain, every slice was dropped — including the attack="none" ones we did cover — which is why the result came back with scores: [].

We have now re-scored the complete test.csv and updated leaderboard/submissions/plagiat-ai-detector/predictions.json on this branch:

• 672,000 records, exactly one score per id, no nulls, all scores in [0, 1]
• the id set was checked against a canonical index built from the official test.csv (1,221,088,550 bytes, sha256 96867bb0cebf4c49…): 0 missing, 0 extra, 0 duplicates
• predictions.json sha256 81381eb424a599129bfcdf34ccf9630371f8609ae47fd487123750607f596cb3, 44,949,374 bytes
• detector unchanged from the original submission: GeorgeDrayson/modernbert-ai-detection-raid-mage @ e047d9c5, batch size 1
• metadata.json is unchanged — the diff is this one file

Could you re-approve the evaluation workflow when you have a moment? Happy to fix anything else you spot. Thanks for the benchmark and for the patience.

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

Plagiat AI Detector

Release date: 2026-07-26

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 97.64 and a TPR of 94.19% at FPR=5% and 88.32% at FPR=1%.
Without adversarial attacks, it achieved AUROC of 99.12 and a TPR of 99.12% at FPR=5% and 98.20% 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!

[cristeamarinescu-cpu]

Thank you, @liamdugan — The bot found our new relese an the evaluation ran cleanly this time. The results are exactly what we needed. We appreciate the benchmark. Best regards!