family,iterations,accuracy,macro_f1,handled_correctly,automation_rate,auto_precision,misroute_rate,unnecessary_escalation_rate,invalid_labels,errors,latency_p50_s,latency_p95_s,throughput_per_s,eval_wall_s,total_wall_s,cost_per_1k_usd,total_cost_usd,cost_per_correct_call_usd,ops_cost_per_1k_usd,cached_input_share,cache_write_tokens,cache_setup_cost_usd,cost_per_1m_calls_usd claude-haiku-5-5_explicit,2,0.608,0.56202,0.516,0.598,0.84213,0.096,0.388,0,0,1.1854,1.7485,4.615,108.48,136.04,0.12187,0.06094,0.00024,2970.12187,0.96464,0,0.0,121.873 jev,2,0.636,0.57125,0.476,0.55,0.84722,0.084,0.44,0,0,0.18773,0.27493,29.82,16.77,20.67,0.17009,0.08505,0.00036,3090.17009,0.0,0,0.0,170.0935 gemini-3.1-flash-lite_explicit,2,0.62,0.55509,0.472,0.558,0.8354,0.1,0.426,0,0,0.95342,1.41559,5.87,85.16,103.98,0.17586,0.08793,0.00038,3210.17586,0.94196,0,0.00132,175.8635 gpt-6-luna_explicit,2,0.652,0.59852,0.402,0.486,0.80304,0.096,0.502,0,0,2.67106,5.89073,1.86,269.21,331.25,0.10302,0.05151,0.00026,3530.10302,0.95156,0,0.0,103.02 claude-sonnet-5-5_explicit,2,0.644,0.58277,0.348,0.384,0.87712,0.048,0.604,0,0,1.42644,2.13155,3.87,129.3,160.98,2.3701,1.18505,0.00681,3562.3701,0.96464,0,0.0,2370.096 classic_tfidf_lr,2,0.304,0.245,0.104,0.138,0.6487,0.048,0.842,0,0,3e-05,3e-05,31043.82,0.02,4.41,0.0,0.0,0.0,4790.0,,0,0.0,0.0