eval-harness

metric

untyped_f1

better
higher
kind
quality: measured against a reference or by a judge
decides
no goal
weighed by
no goal
recorded on
few_nerd_280

01what it measures, and how

F1 on the entity texts alone, types ignored: did the arm find the entity at all.

A project explains its own metric at length with METRIC_DESCRIPTIONS = {"untyped_f1": {"summary": "…", "details": "…"}}.

02on few_nerd_280

gold/few_nerd_280fn_span_marker_n200_v1: 0.837fn_anthropic_l_n200_v1: 0.782fn_openai_m_n200_v1: 0.778fn_gemma_m_n200_v1: 0.775fn_openai_l_n200_v1: 0.769fn_gemma_l_n200_v1: 0.769fn_anthropic_m_n200_v1: 0.767fn_qwen_s_n200_v1: 0.765fn_qwen_m_n200_v1: 0.753fn_qwen_l_n200_v1: 0.728fn_anthropic_s_n200_v1: 0.728fn_deepseek_m_n200_v1: 0.727fn_glm_m_n200_v1: 0.715fn_llama_l_n200_v1: 0.715fn_deepseek_s_n200_v1: 0.7fn_mistral_l_n200_v1: 0.695fn_deepseek_l_n200_v1: 0.694fn_mistral_m_n200_v1: 0.687fn_glm_s_n200_v1: 0.686fn_openai_s_n200_v1: 0.685fn_llama_s_n200_v1: 0.681fn_llama_m_n200_v1: 0.677fn_gemma_s_n200_v1: 0.676fn_glm_l_n200_v1: 0.653fn_mistral_s_n200_v1: 0.645fn_capitalized_n200_v1: 0.619fn_gliner_n200_v1: 0.552fn_nothing_n200_v1: 0.1250.125better · 0.837

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