[woldy]

Detector: decting-v0.13

Method: Qwen3-0.6B + LoRA (r=16) sequence classification, bf16. Successor to decting-v0.10 (currently on the leaderboard) — from char-TF-IDF to a fine-tuned 0.6B model.

Training: RAID train split targeted sampling (25,968 rows — human with wiki/reddit domain boost, base-model floors for gpt2/mpt/mistral/cohere/gpt3, sampling-decoding priority, 30% synonym-attack augmentation; title-grouped, no test leakage) + 2,945 rows of 2026-generation-model text (Kimi K2.5 / GLM-5.3 / DeepSeek-V4.1 / GPT-6-Sol outputs from public corpora).

Validation beyond RAID:

• Stress-tested on unseen 2025 flagships (arena data, 12 families x 300): AUROC 0.998
• Blind test on unseen 2026 flagships (5 generators incl. GPT-6-Sol): AUROC 0.9999, 100% detection at 0.5
• v0.10 TF-IDF baseline on the same sets: 0.49-0.53 (random) — the upgrade is qualitative

Files:

• leaderboard/submissions/decting-v0.13/predictions.json — full test set (672,000 rows)
• leaderboard/submissions/decting-v0.13/metadata.json

Thanks for maintaining RAID!

[github-actions]

Eval run succeeded! Link to run: link

Here are the results of the submission(s):

decting-v0.13

Release date: 2026-10-08

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

Warning

Failed to find threshold values that achieve False Positive Rate(s): (['1%']) on all domains. This submission will not appear in the main leaderboard for those FPR values; it will only be visible within the splits in which the target FPR was achieved.
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 92.11 and a TPR of 96.89% at FPR=5%.
Without adversarial attacks, it achieved AUROC of 93.06 and a TPR of 98.80% at FPR=5%.

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]

Hi @woldy please let me know if you'd like me to merge this to the leaderboard

[woldy]

Yes, please merge. Thanks! We'll follow up with an attack-augmented v0.14 once it's validated.