eval-harness

metric

rho

better
higher
kind
quality: measured against a reference or by a judge
decides
no goal
weighed by
no goal
recorded on
summeval_100

01what it measures, and how

Spearman's rho between the judge's scores and the experts' ratings of an article's 16 summaries, averaged over the four dimensions: kept for comparison with the literature.

A project explains its own metric at length with METRIC_DESCRIPTIONS = {"rho": {"summary": "…", "details": "…"}}.

02on summeval_100

gold/summeval_100se_gemma_l_pubtext_n100_v1 · pair: 0.611se_openai_l_n100_v1: 0.602se_gemma_l_pubtextcw_n100_v1 · pair: 0.598se_gemma_l_n100_v1: 0.598se_openai_m_n100_v1: 0.587se_anthropic_l_n100_v1: 0.587se_qwen37max_n100_v1: 0.581se_qwen_s_n100_v1: 0.581se_glm_m_n100_v1: 0.571se_qwen_m_n100_v1: 0.571se_glm_l_n100_v1: 0.546se_deepseek_m_n100_v1: 0.542se_anthropic_s_n100_v1: 0.541se_anthropic_m_n100_v1: 0.539se_gemma_m_n100_v1: 0.532se_mistral_m_n100_v1: 0.531se_deepseek_l_n100_v1: 0.53se_mistral_l_n100_v1: 0.514se_deepseek_s_n100_v1: 0.514se_llama_l_n100_v1: 0.508se_glm_s_n100_v1: 0.502se_mistral_s_n100_v1: 0.476se_unievalpub_n100_v1: 0.467se_gemma_s_n100_v1: 0.462se_openai_s_n100_v1: 0.461se_llama_m_n100_v1: 0.455se_unieval_n100_v1: 0.439se_llama_s_n100_v1: 0.398se_bartscore_n100_v1: 0.325se_rouge2_n100_v1: 0.201se_rouge1_n100_v1: 0.19se_rougeL_n100_v1: 0.184se_coverage_n100_v1: 0.13se_length_n100_v1: 0.048se_random_n100_v1: 0.00290.0029better · 0.611

scored against silverscored by a judgewrote the reference it is scored againstscored against a label set, not a reference setunmarked rows are scored against gold

armmean rhoitemsscored against
se_gemma_l_pubtext_n100_v1 · pair0.611100gold/summeval_100 gold
se_openai_l_n100_v10.602100gold/summeval_100 gold
se_gemma_l_pubtextcw_n100_v1 · pair0.598100gold/summeval_100 gold
se_gemma_l_n100_v10.598100gold/summeval_100 gold
se_openai_m_n100_v10.587100gold/summeval_100 gold
se_anthropic_l_n100_v10.587100gold/summeval_100 gold
se_qwen37max_n100_v10.581100gold/summeval_100 gold
se_qwen_s_n100_v10.581100gold/summeval_100 gold
se_glm_m_n100_v10.571100gold/summeval_100 gold
se_qwen_m_n100_v10.571100gold/summeval_100 gold
se_glm_l_n100_v10.546100gold/summeval_100 gold
se_deepseek_m_n100_v10.542100gold/summeval_100 gold
se_anthropic_s_n100_v10.541100gold/summeval_100 gold
se_anthropic_m_n100_v10.539100gold/summeval_100 gold
se_gemma_m_n100_v10.532100gold/summeval_100 gold
se_mistral_m_n100_v10.531100gold/summeval_100 gold
se_deepseek_l_n100_v10.530100gold/summeval_100 gold
se_mistral_l_n100_v10.514100gold/summeval_100 gold
se_deepseek_s_n100_v10.514100gold/summeval_100 gold
se_llama_l_n100_v10.508100gold/summeval_100 gold
se_glm_s_n100_v10.502100gold/summeval_100 gold
se_mistral_s_n100_v10.476100gold/summeval_100 gold
se_unievalpub_n100_v10.467100gold/summeval_100 gold
se_gemma_s_n100_v10.462100gold/summeval_100 gold
se_openai_s_n100_v10.461100gold/summeval_100 gold
se_llama_m_n100_v10.455100gold/summeval_100 gold
se_unieval_n100_v10.439100gold/summeval_100 gold
se_llama_s_n100_v10.398100gold/summeval_100 gold
se_bartscore_n100_v10.325100gold/summeval_100 gold
se_rouge2_n100_v10.201100gold/summeval_100 gold
se_rouge1_n100_v10.190100gold/summeval_100 gold
se_rougeL_n100_v10.184100gold/summeval_100 gold
se_coverage_n100_v10.130100gold/summeval_100 gold
se_length_n100_v10.048100gold/summeval_100 gold
se_random_n100_v10.003100gold/summeval_100 gold