Text Quality Meta-evaluation on Topical-Chat (Local) benchmark leaderboard · SOTA2 Research

Text Quality Meta-evaluation on Topical-Chat (Local)

0.831UnderstandabilityGPT-4o mini

Evaluation Results

Method	Links
GPT-4o mini2025.02		0.831	0.717	0.783	0.683	0.621	0.727
CompassJudger-32B2025.02		0.795	0.679	0.783	0.652	0.586	0.699
Qwen-2.5-72B2025.02		0.781	0.61	0.821	0.633	0.612	0.691
Gemma-2-27B2025.02		0.781	0.745	0.66	0.705	0.531	0.684
GPT-4o2025.02		0.776	0.714	0.807	0.702	0.595	0.719
Phi-4-14B2025.02		0.774	0.624	0.726	0.643	0.55	0.663
GPT-4 Turbo2025.02		0.764	0.721	0.776	0.64	0.624	0.705
DeepSeek-V32025.02		0.75	0.731	0.814	0.681	0.662	0.728
Llama-3.1-70B2025.02		0.745	0.695	0.757	0.719	0.571	0.698
GPT-3.5 Turbo2025.02		0.7	0.617	0.698	0.69	0.555	0.652
Prometheus-2-8x7B2025.02		0.695	0.612	0.633	0.602	0.552	0.619
Prometheus-2-7B2025.02		0.688	0.6	0.669	0.621	0.593	0.634
CRITIQUELLM-6B2025.02		0.664	0.586	0.636	0.471	0.564	0.584
Themis-8B2025.02		0.588	0.462	0.543	0.443	0.357	0.479
Auto-J-13B2025.02		0.533	0.445	0.533	0.462	0.448	0.484
Prometheus-13B2025.02		0.433	0.54	0.343	0.557	0.49	0.473