eval-harness

metric

untyped_tp

better
neutral: a description
kind
descriptive: says what the output was, never whether it was good
decides
no goal
weighed by
no goal
recorded on
few_nerd_280

01what it measures, and how

True positives on texts alone, types ignored.

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

02on few_nerd_280

gold/few_nerd_280fn_span_marker_n200_v1: 2.375fn_anthropic_l_n200_v1: 2.3fn_gemma_s_n200_v1: 2.3fn_openai_l_n200_v1: 2.289fn_deepseek_l_n200_v1: 2.282fn_deepseek_m_n200_v1: 2.279fn_gemma_l_n200_v1: 2.268fn_qwen_m_n200_v1: 2.268fn_openai_m_n200_v1: 2.264fn_gemma_m_n200_v1: 2.254fn_deepseek_s_n200_v1: 2.25fn_glm_m_n200_v1: 2.25fn_qwen_s_n200_v1: 2.243fn_llama_l_n200_v1: 2.232fn_anthropic_s_n200_v1: 2.221fn_openai_s_n200_v1: 2.211fn_mistral_l_n200_v1: 2.189fn_mistral_s_n200_v1: 2.189fn_llama_m_n200_v1: 2.182fn_anthropic_m_n200_v1: 2.179fn_llama_s_n200_v1: 2.15fn_mistral_m_n200_v1: 2.146fn_qwen_l_n200_v1: 2.114fn_capitalized_n200_v1: 2.107fn_glm_s_n200_v1: 2.039fn_glm_l_n200_v1: 1.975fn_gliner_n200_v1: 1.829fn_nothing_n200_v1: 00better · 2.375

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