LLM-as-a-Judge on JudgeLM (test)
79.59AgreementQwen2.5-72B
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen2.5-72BCapacity=72B2026.02 | 79.59 | 71.31 | 60.95 | 61.92 | |
| FairJudge-8BCapacity=8B, Backbone=Qwen3-VL2026.02 | 78.82 | 66.77 | 66.93 | 66.78 | |
| JudgeLMCapacity=7B, Specialized Judge=true2026.02 | 78 | 66.02 | 67.61 | 66.66 | |
| FlexVLCapacity=7B, Specialized Judge=true2026.02 | 77.44 | 64.82 | 57.49 | 57.42 | |
| FairJudge-4BCapacity=4B, Backbone=Qwen3-VL2026.02 | 75.07 | 71.03 | 59.59 | 60.95 | |
| GLM-4-9BCapacity=9B2026.02 | 72.42 | 59.06 | 53.31 | 52.54 | |
| Qwen3-VL-8BCapacity=8B2026.02 | 72.15 | 58.77 | 58.78 | 58.72 | |
| DeepSeek-V3-671BCapacity=671B2026.02 | 72.04 | 59.38 | 60.37 | 59.75 | |
| InternVL3-14BCapacity=14B2026.02 | 71.96 | 62.26 | 64.69 | 61.23 | |
| FairJudge-2BCapacity=2B, Backbone=Qwen3-VL2026.02 | 71.2 | 59.88 | 52.36 | 51.48 | |
| PandaLMCapacity=7B, Specialized Judge=true2026.02 | 67.42 | 44.99 | 48.53 | 46.66 | |
| LLaVA-1.5-13BCapacity=13B2026.02 | 48.59 | 41.61 | 40.45 | 39.69 | |
| LLaVA-1.5-7BCapacity=7B2026.02 | 40.8 | 43.11 | 41.16 | 34.34 |