{ "n_tasks": 27, "single": { "deepseek-v4-flash": { "pass": 0.7407407407407407, "score": 0.7194407407407408, "cost": 0.025244660000000002, "latency_p50": 32.60991859436035 }, "gpt-5.4-nano": { "pass": 0.7777777777777778, "score": 0.7777111111111111, "cost": 0.030138250000000002, "latency_p50": 6.396074295043945 }, "gpt-5.4-mini": { "pass": 0.8518518518518519, "score": 0.8517851851851852, "cost": 0.42622575, "latency_p50": 27.368282556533813 }, "gemini-3.5-flash": { "pass": 0.4444444444444444, "score": 0.4444444444444444, "cost": 0.8562555000000001, "latency_p50": 17.11278009414673 }, "gpt-5.5": { "pass": 0.7777777777777778, "score": 0.7777111111111111, "cost": 2.547895, "latency_p50": 27.903853178024292 } }, "router": { "0.4": { "pass": 0.7407407407407407, "score": 0.7194407407407408, "cost": 0.02524466 }, "0.5": { "pass": 0.7777777777777778, "score": 0.7564777777777778, "cost": 0.025137299999999998 }, "0.6": { "pass": 0.7407407407407407, "score": 0.7406740740740742, "cost": 0.03005728 }, "0.7": { "pass": 0.7777777777777778, "score": 0.7777111111111111, "cost": 0.14011191 }, "0.8": { "pass": 0.7777777777777778, "score": 0.7777111111111111, "cost": 0.25106635 }, "0.9": { "pass": 0.7777777777777778, "score": 0.7777111111111111, "cost": 0.25106635 } }, "cascade": { "price_order": { "pass": 1.0, "score": 0.9787, "cost": 0.11071721000000001 }, "router_start_0.5": { "pass": 1.0, "score": 0.9787, "cost": 0.10954415 }, "router_start_0.6": { "pass": 0.9629629629629629, "score": 0.9628962962962964, "cost": 0.67018278 }, "router_start_0.7": { "pass": 0.9259259259259259, "score": 0.9258592592592593, "cost": 0.93017776 } }, "k_ablation": { "1": { "pass": 0.7777777777777778, "score": 0.7777111111111111, "cost": 0.03041265 }, "2": { "pass": 0.7407407407407407, "score": 0.7406740740740742, "cost": 0.2955786 }, "4": { "pass": 0.7037037037037037, "score": 0.7036370370370371, "cost": 0.06224015000000001 }, "8": { "pass": 0.7407407407407407, "score": 0.7406740740740742, "cost": 0.03005728 }, "12": { "pass": 0.7407407407407407, "score": 0.7194407407407408, "cost": 0.024380539999999996 }, "16": { "pass": 0.7407407407407407, "score": 0.7194407407407408, "cost": 0.02590911 } }, "backend": { "pybed": { "pass": 0.7407407407407407, "score": 0.7406740740740742, "cost": 0.03005728 }, "hash": { "pass": 0.8518518518518519, "score": 0.8305518518518519, "cost": 0.02901353 } }, "oracle": { "pass": 1.0, "score": 1.0, "cost": 0.09019056000000003 } }