Text Classification on Claude9 (test)
57.5Macro-F1Human
Evaluation results
| Method | Links | |
|---|---|---|
| HumanDownstream Model=Llama, CFT=true2026.06 | 57.5 | |
| Human Annotators2026.06 | 55.51 | |
| HumanDownstream Model=Qwen, CFT=false2026.06 | 52.03 | |
| HumanDownstream Model=Llama, CFT=false2026.06 | 51.9 | |
| EvoPoolDownstream Model=Qwen, CFT=true2026.06 | 45 | |
| HumanDownstream Model=Qwen, CFT=true2026.06 | 43.58 | |
| EvoPoolDownstream Model=Llama, CFT=true2026.06 | 42.59 | |
| EvoPoolDownstream Model=Llama, CFT=false2026.06 | 39.08 | |
| EvoPoolDownstream Model=Qwen, CFT=false2026.06 | 38.27 | |
| HumanDownstream Model=RoB, CFT=true2026.06 | 37.73 | |
| EvoPoolBackbone=gpt-4o-mini2026.06 | 36.45 | |
| EvoPoolDownstream Model=RoB, CFT=true2026.06 | 35.28 | |
| EvoPoolDownstream Model=RoB, CFT=false2026.06 | 34.13 | |
| HumanDownstream Model=RoB, CFT=false2026.06 | 19.45 |