== L: lem:lines
  400 random (u, s, a, b, y), exact: phi == max over monotone sigma: True
== P: lem:patch
  300 random points of patch cells: lower and upper hull values bracket g(a): True
== C: cells (single-polynomial and vertex-decoupled), lemma chain on random configurations
  222 configurations {'cell10f_3': 127, 'cell10d': 21, 'cell10f': 18, 'cell10d_3': 56}: E|L| <= sum_v w_v E P_v(M), E P_v(M) <= sum c G + F_v < T on all: True; max E|L|/T = 0.989485
== N: near-sphere boxes on random configurations
  245 configurations: E|L| <= S3: True; max E|L|/S3 = 0.999530
== K: comparison boxes on random configurations
  200 configurations: P >= R on the grid, P(-v) = 0, P(u) = l_R(u), E|L| = E|X| + E R(M) <= E|X| + E P(M) <= E|X + b chi_4| <= T: True; max E|L|/T = 0.999172
== A: adversarial search (hill climbing on E|L| / |alpha|)
  u = 2.6000: max found E|L|/T = 0.999999999; top |alpha_i| [0.5 0.5 0.5 0.5]
  u = 2.6050: max found E|L|/T = 0.999999999; top |alpha_i| [0.5 0.5 0.5 0.5]
  u = 2.6100: max found E|L|/T = 0.999999996; top |alpha_i| [0.5001 0.5    0.5    0.5   ]
  u = 2.6150: max found E|L|/T = 0.999999994; top |alpha_i| [0.5001 0.5    0.5    0.4999]
  u = 2.6200: max found E|L|/T = 1.000000000; top |alpha_i| [0.5 0.5 0.5 0.5]
  no configuration above T: True
summary {'L': True, 'P': True, 'C': True, 'N': True, 'K': True, 'A': np.True_}; 11 s
ALL SEMANTIC TESTS PASS

real	0m11.766s
user	0m12.381s
sys	0m0.056s
