Refresh leaderboard: 7 V2 + 6 V1 rows with reward_rate (judge-graded)

leaderboard/results.csv

@@ -0,0 +1,14 @@
+model,harness,dataset,passed,total,pass_rate,reward_rate,wall_hours
+glm-5.1,hermes,v2,24,130,48.46,18.46,
+claude-opus-4-7,hermes,v2,10,75,54.67,13.33,
+gpt-5.5,hermes,v2,9,81,48.15,11.11,
+deepseek-v4-pro,hermes,v2,13,130,43.85,10.00,
+openrouter-owl-alpha,hermes,v2,6,130,14.62,4.62,
+deepseek-v4-flash,hermes,v2,2,130,3.08,1.54,
+glm-5.1,openclaw,v2,0,130,0.00,0.00,
+claude-opus-4-6,hermes,v1,94,153,61.44,61.44,
+claude-sonnet-4-6,hermes,v1,87,153,56.86,56.86,
+claude-haiku-4-5-20251001,hermes,v1,46,153,30.07,30.07,
+gpt-5.4-2026-03-05,hermes,v1,39,153,25.49,25.49,
+gpt-5.4-mini-2026-03-17,hermes,v1,38,153,24.84,24.84,
+kimi-k2.5,hermes,v1,27,153,17.65,17.65,