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

Text Quality Meta-evaluation on SummEval (Local)

0.687CoherenceDeepSeek-V3

Evaluation Results

Method	Links
DeepSeek-V32025.02		0.687	0.663	0.669	0.63	0.662
GPT-4 Turbo2025.02		0.683	0.723	0.643	0.643	0.673
Qwen-2.5-72B2025.02		0.683	0.703	0.61	0.63	0.657
Gemma-2-27B2025.02		0.678	0.668	0.65	0.622	0.655
CompassJudger-32B2025.02		0.665	0.654	0.593	0.626	0.634
GPT-4o2025.02		0.661	0.707	0.661	0.646	0.669
GPT-4o mini2025.02		0.654	0.64	0.654	0.617	0.641
CRITIQUELLM-6B2025.02		0.648	0.642	0.565	0.579	0.609
Llama-3.1-70B2025.02		0.642	0.547	0.625	0.595	0.602
Phi-4-14B2025.02		0.636	0.695	0.659	0.617	0.652
Prometheus-2-8x7B2025.02		0.623	0.621	0.505	0.551	0.575
GPT-3.5 Turbo2025.02		0.609	0.615	0.569	0.601	0.598
Prometheus-2-7B2025.02		0.597	0.531	0.517	0.539	0.546
Auto-J-13B2025.02		0.565	0.574	0.496	0.541	0.544
Prometheus-13B2025.02		0.476	0.49	0.522	0.49	0.495
Themis-8B2025.02		0.368	0.385	0.307	0.361	0.355