Junk Data Detection on Brain Rot junk data detection M2 (test)
0.8727Top-5 auPRCRISE
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| RISEModel=OLMo-3-7B, Top-K selection threshold=1,000, lambda_rh=0.7, lambda_gh=1.0, Kr=256, Kh=192, Kg=1282026.04 | 0.8727 | 0.9516 | 0.8586 | 0.9456 | 0.8371 | 0.9322 | 0.8252 | 0.9265 | |
| RISEModel=OLMo-3-32B, Top-K selection threshold=1,000, lambda_rh=0.7, lambda_gh=1.0, Kr=256, Kh=192, Kg=1282026.04 | 0.8637 | 0.9663 | 0.8501 | 0.9571 | 0.8333 | 0.9328 | 0.8145 | 0.9228 | |
| RISEModel=OLMo-2-1B, Top-K selection threshold=1,000, lambda_rh=0.7, lambda_gh=1.0, Kr=256, Kh=192, Kg=1282026.04 | 0.8603 | 0.9574 | 0.8489 | 0.9404 | 0.8244 | 0.9289 | 0.807 | 0.9228 | |
| RapidInModel=OLMo-2-1B, Top-K selection threshold=1,0002026.04 | 0.8583 | 0.8766 | 0.8501 | 0.8929 | 0.827 | 0.8989 | 0.8077 | 0.894 | |
| RapidInModel=OLMo-3-7B, Top-K selection threshold=1,0002026.04 | 0.8328 | 0.8414 | 0.8314 | 0.8686 | 0.8202 | 0.8874 | 0.7965 | 0.8815 | |
| Embedding Similarity (E5-base-v2)Top-K selection threshold=1,0002026.04 | 0.7048 | 0.7825 | 0.6226 | 0.7474 | 0.5468 | 0.7231 | 0.4852 | 0.695 | |
| TrackStarModel=OLMo-3-7B, Top-K selection threshold=1,0002026.04 | 0.6966 | 0.5759 | 0.6665 | 0.5956 | 0.62 | 0.6064 | 0.5817 | 0.6133 | |
| TrackStarModel=OLMo-3-32B, Top-K selection threshold=1,0002026.04 | 0.6941 | 0.4306 | 0.7019 | 0.4919 | 0.696 | 0.5547 | 0.6864 | 0.5948 | |
| BM25Top-K selection threshold=1,0002026.04 | 0.5627 | 0.6307 | 0.5336 | 0.626 | 0.5007 | 0.6739 | 0.4747 | 0.6748 | |
| TrackStarModel=OLMo-2-1B, Top-K selection threshold=1,0002026.04 | 0.5244 | 0.5938 | 0.4761 | 0.5849 | 0.4238 | 0.5912 | 0.3917 | 0.5938 | |
| ZO-InfModel=OLMo-3-7B, Top-K selection threshold=1,0002026.04 | 0.4997 | 0.7228 | 0.3685 | 0.6977 | 0.2803 | 0.6791 | 0.2426 | 0.6726 | |
| ZO-InfModel=OLMo-2-1B, Top-K selection threshold=1,0002026.04 | 0.4515 | 0.6225 | 0.3434 | 0.6248 | 0.3073 | 0.6489 | 0.2898 | 0.6558 |