Text Classification on Trec Merge Imbalance pool split
93.1Balanced Acc (10% cost)CORESET
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CORESETBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 93.1 | 97.89 | |
| PriorALBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 90.91 | 97.96 | |
| BADGEBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 90.58 | 97.16 | |
| ConfidenceBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 89.57 | 97.3 | |
| MarginBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 89.48 | 95.91 | |
| GALAXYBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 89.19 | 94.27 | |
| RandomBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 82.39 | 88.78 | |
| BASEBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 78.69 | 90.6 | |
| Most LikelyBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 69.99 | 82.37 | |
| DIRECTBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 64.88 | 67.32 | |
| Pretrained PriorsBackbone=RoBERTa-Base, Foundation Model=Llama-3.1-8B2026.05 | 41.03 | 41.03 |