OUTPUT of run 2 of at most 100:
```
WITHIN-LAYER QK COSINES, THEN OV COSINES
L 0 QK
[[ 1.     0.012  0.007  0.137  0.013  0.044]
 [ 0.012  1.     0.69  -0.026  0.57   0.026]
 [ 0.007  0.69   1.    -0.063  0.492 -0.075]
 [ 0.137 -0.026 -0.063  1.    -0.044  0.01 ]
 [ 0.013  0.57   0.492 -0.044  1.     0.033]
 [ 0.044  0.026 -0.075  0.01   0.033  1.   ]]
OV
[[ 1.     0.018  0.017 -0.035  0.026  0.018]
 [ 0.018  1.     0.925  0.006  0.868  0.643]
 [ 0.017  0.925  1.     0.009  0.81   0.692]
 [-0.035  0.006  0.009  1.     0.011 -0.001]
 [ 0.026  0.868  0.81   0.011  1.     0.592]
 [ 0.018  0.643  0.692 -0.001  0.592  1.   ]]
L 1 QK
[[1.    0.358 0.373 0.346 0.334 0.377]
 [0.358 1.    0.372 0.362 0.302 0.399]
 [0.373 0.372 1.    0.368 0.375 0.352]
 [0.346 0.362 0.368 1.    0.32  0.396]
 [0.334 0.302 0.375 0.32  1.    0.289]
 [0.377 0.399 0.352 0.396 0.289 1.   ]]
OV
[[ 1.    -0.02   0.03  -0.012  0.001  0.029]
 [-0.02   1.    -0.009 -0.028 -0.018  0.016]
 [ 0.03  -0.009  1.     0.001  0.014 -0.017]
 [-0.012 -0.028  0.001  1.    -0.001 -0.015]
 [ 0.001 -0.018  0.014 -0.001  1.    -0.014]
 [ 0.029  0.016 -0.017 -0.015 -0.014  1.   ]]
L 2 QK
[[1.    0.129 0.161 0.117 0.18  0.135]
 [0.129 1.    0.116 0.086 0.121 0.117]
 [0.161 0.116 1.    0.101 0.145 0.124]
 [0.117 0.086 0.101 1.    0.093 0.093]
 [0.18  0.121 0.145 0.093 1.    0.134]
 [0.135 0.117 0.124 0.093 0.134 1.   ]]
OV
[[ 1.    -0.023 -0.039 -0.022 -0.005  0.001]
 [-0.023  1.    -0.001  0.002 -0.012 -0.008]
 [-0.039 -0.001  1.     0.013  0.004  0.022]
 [-0.022  0.002  0.013  1.     0.001  0.   ]
 [-0.005 -0.012  0.004  0.001  1.     0.004]
 [ 0.001 -0.008  0.022  0.     0.004  1.   ]]
L 3 QK
[[1.    0.058 0.083 0.06  0.007 0.108]
 [0.058 1.    0.07  0.034 0.042 0.075]
 [0.083 0.07  1.    0.031 0.036 0.094]
 [0.06  0.034 0.031 1.    0.037 0.058]
 [0.007 0.042 0.036 0.037 1.    0.028]
 [0.108 0.075 0.094 0.058 0.028 1.   ]]
OV
[[ 1.    -0.012  0.024  0.011 -0.002  0.034]
 [-0.012  1.    -0.021  0.024  0.059  0.001]
 [ 0.024 -0.021  1.     0.003  0.077  0.026]
 [ 0.011  0.024  0.003  1.     0.042 -0.002]
 [-0.002  0.059  0.077  0.042  1.     0.057]
 [ 0.034  0.001  0.026 -0.002  0.057  1.   ]]

L0 WEIGHT-ONLY EIGENSPECTRA AND ALIGNMENTS
H0 raw: eig_top12=[ 0.113+0.216j  0.113-0.216j -0.166+0.039j -0.166-0.039j -0.141+0.034j -0.141-0.034j -0.11 +0.061j -0.11 -0.061j -0.09 +0.068j -0.09 -0.068j -0.106+0.j    -0.092+0.009j], sv_top8=[0.853 0.613 0.368 0.287 0.279 0.244 0.232 0.21 ]
H0 gain: eig_top12=[ 0.079+0.j     0.049+0.j    -0.024+0.006j -0.024-0.006j -0.023+0.j     0.018+0.013j  0.018-0.013j -0.02 +0.j    -0.014+0.012j -0.014-0.012j -0.014+0.003j -0.014-0.003j], sv_top8=[0.248 0.102 0.047 0.043 0.037 0.036 0.034 0.031]
top left/right cos [ 0.488 -0.117  0.501 -0.2   -0.489]
top left/right coordinate concentration [0.094 0.147 0.063 0.067 0.065] [0.103 0.085 0.084 0.088 0.08 ]
H1 raw: eig_top12=[-0.334+0.j    -0.225+0.j    -0.067+0.012j -0.067-0.012j -0.048+0.j     0.034+0.j    -0.027+0.j    -0.023+0.j    -0.01 +0.004j -0.01 -0.004j -0.009+0.002j -0.009-0.002j], sv_top8=[0.838 0.311 0.219 0.125 0.091 0.086 0.066 0.039]
H1 gain: eig_top12=[-0.062+0.j    -0.01 +0.j    -0.009+0.j     0.004+0.j    -0.004+0.j    -0.003+0.002j -0.003-0.002j -0.002-0.j    -0.002+0.j    -0.002+0.001j -0.002-0.001j -0.001+0.j   ], sv_top8=[0.15  0.042 0.021 0.012 0.01  0.007 0.007 0.004]
top left/right cos [-0.406 -0.111 -0.405 -0.516 -0.422]
top left/right coordinate concentration [0.138 0.115 0.048 0.05  0.299] [0.15  0.046 0.023 0.048 0.049]
H2 raw: eig_top12=[-0.237+0.j    -0.095+0.029j -0.095-0.029j -0.061+0.j    -0.016+0.011j -0.016-0.011j -0.014+0.j    -0.011+0.005j -0.011-0.005j -0.008-0.j    -0.008+0.j     0.   -0.007j], sv_top8=[0.543 0.183 0.157 0.133 0.072 0.066 0.046 0.046]
H2 gain: eig_top12=[-0.057+0.j    -0.007+0.003j -0.007-0.003j -0.003+0.001j -0.003-0.001j -0.003+0.j    -0.001+0.j    -0.001-0.j    -0.001+0.001j -0.001-0.001j  0.001+0.001j  0.001-0.001j], sv_top8=[0.102 0.018 0.016 0.012 0.008 0.007 0.005 0.004]
top left/right cos [-0.539 -0.388 -0.501 -0.573 -0.215]
top left/right coordinate concentration [0.155 0.077 0.185 0.049 0.061] [0.12  0.044 0.048 0.051 0.057]
H3 raw: eig_top12=[0.724+0.691j 0.724-0.691j 0.446+0.842j 0.446-0.842j 0.572+0.722j 0.572-0.722j 0.642+0.614j 0.642-0.614j 0.676+0.463j 0.676-0.463j 0.109+0.762j 0.109-0.762j], sv_top8=[1.114 1.092 1.07  1.064 1.03  0.992 0.968 0.954]
H3 gain: eig_top12=[0.137+0.133j 0.137-0.133j 0.11 +0.089j 0.11 -0.089j 0.095+0.098j 0.095-0.098j 0.084+0.104j 0.084-0.104j 0.064+0.114j 0.064-0.114j 0.096+0.068j 0.096-0.068j], sv_top8=[0.244 0.223 0.201 0.182 0.179 0.157 0.155 0.149]
top left/right cos [0.658 0.673 0.628 0.553 0.593]
top left/right coordinate concentration [0.096 0.336 0.126 0.116 0.074] [0.317 0.388 0.196 0.174 0.302]
H4 raw: eig_top12=[-0.172+0.j    -0.102+0.013j -0.102-0.013j -0.05 +0.j    -0.034+0.j    -0.018+0.006j -0.018-0.006j -0.009+0.009j -0.009-0.009j -0.009+0.005j -0.009-0.005j  0.01 +0.j   ], sv_top8=[0.51  0.227 0.185 0.124 0.101 0.066 0.065 0.049]
H4 gain: eig_top12=[-0.015+0.j    -0.01 +0.j    -0.008+0.j    -0.003+0.j    -0.003-0.j    -0.002+0.001j -0.002-0.001j -0.   +0.002j -0.   -0.002j -0.001+0.001j -0.001-0.001j -0.002+0.j   ], sv_top8=[0.08  0.022 0.019 0.013 0.011 0.008 0.007 0.006]
top left/right cos [-0.166 -0.48  -0.366 -0.27  -0.135]
top left/right coordinate concentration [0.185 0.21  0.069 0.062 0.075] [0.079 0.1   0.047 0.059 0.073]
H5 raw: eig_top12=[-0.028+0.j     0.019+0.j    -0.008+0.j    -0.005+0.005j -0.005-0.005j -0.005+0.002j -0.005-0.002j -0.004+0.004j -0.004-0.004j -0.005+0.002j -0.005-0.002j  0.   +0.005j], sv_top8=[0.18  0.048 0.024 0.02  0.02  0.019 0.019 0.018]
H5 gain: eig_top12=[-0.005+0.j     0.001+0.j    -0.001+0.j    -0.   +0.001j -0.   -0.001j -0.001+0.j    -0.001+0.j    -0.001-0.j     0.   +0.001j  0.   -0.001j -0.   +0.001j -0.   -0.001j], sv_top8=[0.032 0.004 0.003 0.003 0.002 0.002 0.002 0.002]
top left/right cos [-0.145  0.202 -0.338 -0.096 -0.131]
top left/right coordinate concentration [0.121 0.049 0.111 0.07  0.074] [0.049 0.045 0.187 0.067 0.074]

L0 QK TOP-DIRECTION COSINES (signed arbitrary; absolute values shown)
left [[1.    0.138 0.075 0.626 0.089 0.912]
 [0.138 1.    0.887 0.142 0.914 0.067]
 [0.075 0.887 1.    0.453 0.869 0.18 ]
 [0.626 0.142 0.453 1.    0.185 0.662]
 [0.089 0.914 0.869 0.185 1.    0.06 ]
 [0.912 0.067 0.18  0.662 0.06  1.   ]]
right [[1.    0.053 0.256 0.533 0.302 0.043]
 [0.053 1.    0.806 0.047 0.686 0.324]
 [0.256 0.806 1.    0.304 0.597 0.528]
 [0.533 0.047 0.304 1.    0.065 0.106]
 [0.302 0.686 0.597 0.065 1.    0.37 ]
 [0.043 0.324 0.528 0.106 0.37  1.   ]]

ATTENTION AFFINE PARAMETER NORMS (no activations)
L0H0: ||qD||=2.753 ||kD||=3.026 ||qb||=0.504 ||kb||=0.725 ||key_linear||=0.043 ||query_linear||=0.106
L0H1: ||qD||=1.904 ||kD||=1.535 ||qb||=0.228 ||kb||=0.261 ||key_linear||=0.013 ||query_linear||=0.027
L0H2: ||qD||=1.628 ||kD||=1.389 ||qb||=0.146 ||kb||=0.314 ||key_linear||=0.007 ||query_linear||=0.033
L0H3: ||qD||=5.150 ||kD||=6.350 ||qb||=0.467 ||kb||=0.543 ||key_linear||=0.043 ||query_linear||=0.052
L0H4: ||qD||=1.611 ||kD||=1.432 ||qb||=0.270 ||kb||=0.550 ||key_linear||=0.017 ||query_linear||=0.069
L0H5: ||qD||=1.080 ||kD||=0.938 ||qb||=0.143 ||kb||=0.171 ||key_linear||=0.005 ||query_linear||=0.015
L1H0: ||qD||=8.551 ||kD||=8.825 ||qb||=6.688 ||kb||=13.795 ||key_linear||=3.975 ||query_linear||=2.983
L1H1: ||qD||=8.267 ||kD||=8.591 ||qb||=6.079 ||kb||=13.929 ||key_linear||=3.689 ||query_linear||=2.828
L1H2: ||qD||=8.040 ||kD||=8.321 ||qb||=5.564 ||kb||=13.162 ||key_linear||=3.052 ||query_linear||=2.523
L1H3: ||qD||=8.128 ||kD||=8.415 ||qb||=5.665 ||kb||=13.742 ||key_linear||=3.391 ||query_linear||=2.648
L1H4: ||qD||=7.819 ||kD||=7.741 ||qb||=5.347 ||kb||=10.986 ||key_linear||=2.329 ||query_linear||=2.037
L1H5: ||qD||=8.316 ||kD||=8.794 ||qb||=6.224 ||kb||=14.146 ||key_linear||=3.896 ||query_linear||=2.857
L2H0: ||qD||=9.875 ||kD||=9.920 ||qb||=4.265 ||kb||=10.635 ||key_linear||=2.021 ||query_linear||=1.811
L2H1: ||qD||=11.498 ||kD||=11.425 ||qb||=3.787 ||kb||=10.117 ||key_linear||=1.881 ||query_linear||=1.635
L2H2: ||qD||=10.317 ||kD||=10.337 ||qb||=3.784 ||kb||=9.886 ||key_linear||=1.609 ||query_linear||=1.418
L2H3: ||qD||=10.057 ||kD||=9.734 ||qb||=3.082 ||kb||=7.454 ||key_linear||=1.060 ||query_linear||=0.841
L2H4: ||qD||=11.573 ||kD||=11.405 ||qb||=4.319 ||kb||=11.042 ||key_linear||=2.307 ||query_linear||=1.981
L2H5: ||qD||=11.724 ||kD||=11.635 ||qb||=3.751 ||kb||=10.382 ||key_linear||=2.024 ||query_linear||=1.749
L3H0: ||qD||=11.379 ||kD||=11.100 ||qb||=1.761 ||kb||=3.665 ||key_linear||=0.501 ||query_linear||=0.447
L3H1: ||qD||=8.763 ||kD||=8.475 ||qb||=1.460 ||kb||=3.019 ||key_linear||=0.280 ||query_linear||=0.306
L3H2: ||qD||=9.584 ||kD||=9.387 ||qb||=1.940 ||kb||=3.360 ||key_linear||=0.495 ||query_linear||=0.341
L3H3: ||qD||=10.731 ||kD||=10.630 ||qb||=1.298 ||kb||=1.116 ||key_linear||=0.221 ||query_linear||=0.181
L3H4: ||qD||=6.101 ||kD||=5.901 ||qb||=1.230 ||kb||=1.267 ||key_linear||=0.144 ||query_linear||=0.117
L3H5: ||qD||=10.323 ||kD||=10.032 ||qb||=1.800 ||kb||=3.333 ||key_linear||=0.493 ||query_linear||=0.422

KEY-SIDE COMPOSITION: trace(A_l B_earlier)/||A_l B_earlier||_F
These are matrix-path diagnostics, not observed attention or induction.
L1H0 [(np.float64(0.134), 'L0H2', np.float64(2.261)), (np.float64(0.102), 'L0H3', np.float64(0.399)), (np.float64(0.1), 'L0H5', np.float64(17.379)), (np.float64(0.096), 'L0H4', np.float64(1.016)), (np.float64(0.094), 'L0H1', np.float64(2.843))]
L1H1 [(np.float64(0.216), 'L0H3', np.float64(0.422)), (np.float64(0.065), 'L0H2', np.float64(2.168)), (np.float64(0.041), 'L0H5', np.float64(16.984)), (np.float64(0.018), 'L0H0', np.float64(0.115)), (np.float64(0.017), 'L0H1', np.float64(2.729))]
L1H2 [(np.float64(0.167), 'L0H3', np.float64(0.455)), (np.float64(0.106), 'L0H2', np.float64(2.141)), (np.float64(0.1), 'L0H4', np.float64(0.941)), (np.float64(0.09), 'L0H5', np.float64(17.046)), (np.float64(0.078), 'L0H1', np.float64(2.692))]
L1H3 [(np.float64(0.113), 'L0H2', np.float64(2.176)), (np.float64(0.082), 'L0H5', np.float64(17.098)), (np.float64(0.076), 'L0H3', np.float64(0.429)), (np.float64(0.075), 'L0H4', np.float64(0.971)), (np.float64(0.073), 'L0H1', np.float64(2.741))]
L1H4 [(np.float64(0.08), 'L0H2', np.float64(1.852)), (np.float64(0.078), 'L0H4', np.float64(0.815)), (np.float64(0.074), 'L0H5', np.float64(15.135)), (np.float64(0.046), 'L0H1', np.float64(2.335)), (np.float64(0.02), 'L0H0', np.float64(0.122))]
L1H5 [(np.float64(0.198), 'L0H3', np.float64(0.37)), (np.float64(0.102), 'L0H2', np.float64(2.195)), (np.float64(0.062), 'L0H5', np.float64(16.793)), (np.float64(0.061), 'L0H4', np.float64(0.992)), (np.float64(0.052), 'L0H1', np.float64(2.768))]
L2H0 [(np.float64(0.208), 'L1H3', np.float64(0.907)), (np.float64(0.196), 'L0H2', np.float64(1.477)), (np.float64(0.189), 'L0H5', np.float64(10.681)), (np.float64(0.164), 'L1H1', np.float64(0.787)), (np.float64(0.133), 'L0H1', np.float64(1.866))]
L2H1 [(np.float64(0.197), 'L0H5', np.float64(8.667)), (np.float64(0.179), 'L0H2', np.float64(1.219)), (np.float64(0.116), 'L0H1', np.float64(1.55)), (np.float64(0.08), 'L0H4', np.float64(0.587)), (np.float64(0.035), 'L0H0', np.float64(0.193))]
L2H2 [(np.float64(0.184), 'L0H5', np.float64(8.027)), (np.float64(0.182), 'L0H2', np.float64(1.12)), (np.float64(0.123), 'L0H1', np.float64(1.427)), (np.float64(0.099), 'L0H3', np.float64(0.321)), (np.float64(0.075), 'L1H1', np.float64(0.899))]
L2H3 [(np.float64(0.129), 'L0H2', np.float64(0.859)), (np.float64(0.11), 'L0H5', np.float64(6.748)), (np.float64(0.096), 'L0H1', np.float64(1.089)), (np.float64(0.084), 'L0H4', np.float64(0.387)), (np.float64(0.082), 'L0H3', np.float64(0.328))]
L2H4 [(np.float64(0.186), 'L0H2', np.float64(1.634)), (np.float64(0.18), 'L0H5', np.float64(11.718)), (np.float64(0.159), 'L0H3', np.float64(0.43)), (np.float64(0.128), 'L0H1', np.float64(2.055)), (np.float64(0.109), 'L0H4', np.float64(0.75))]
L2H5 [(np.float64(0.21), 'L0H5', np.float64(9.536)), (np.float64(0.187), 'L0H2', np.float64(1.329)), (np.float64(0.112), 'L0H1', np.float64(1.69)), (np.float64(0.081), 'L0H4', np.float64(0.639)), (np.float64(0.018), 'L0H0', np.float64(0.208))]
L3H0 [(np.float64(0.107), 'L0H2', np.float64(0.735)), (np.float64(0.085), 'L0H5', np.float64(5.722)), (np.float64(0.069), 'L0H1', np.float64(0.946)), (np.float64(0.047), 'L0H4', np.float64(0.351)), (np.float64(-0.049), 'L0H0', np.float64(0.218))]
L3H1 [(np.float64(0.093), 'L0H2', np.float64(0.533)), (np.float64(0.075), 'L0H5', np.float64(4.189)), (np.float64(0.072), 'L1H2', np.float64(0.667)), (np.float64(0.069), 'L0H1', np.float64(0.681)), (np.float64(0.051), 'L1H1', np.float64(0.643))]
L3H2 [(np.float64(0.081), 'L0H2', np.float64(0.644)), (np.float64(0.046), 'L0H1', np.float64(0.835)), (np.float64(0.033), 'L0H3', np.float64(0.259)), (np.float64(0.016), 'L0H5', np.float64(5.116)), (np.float64(0.005), 'L1H2', np.float64(0.771))]
L3H3 [(np.float64(0.922), 'L0H3', np.float64(0.435)), (np.float64(0.806), 'L2H5', np.float64(1.826)), (np.float64(0.562), 'L2H1', np.float64(1.866)), (np.float64(0.514), 'L1H0', np.float64(1.236)), (np.float64(0.472), 'L1H5', np.float64(1.124))]
L3H4 [(np.float64(0.276), 'L2H2', np.float64(0.593)), (np.float64(0.272), 'L2H4', np.float64(0.415)), (np.float64(0.202), 'L2H5', np.float64(0.486)), (np.float64(0.19), 'L0H4', np.float64(0.057)), (np.float64(0.182), 'L2H1', np.float64(0.512))]
L3H5 [(np.float64(0.05), 'L0H2', np.float64(0.738)), (np.float64(0.047), 'L0H5', np.float64(5.814)), (np.float64(0.016), 'L0H1', np.float64(0.945)), (np.float64(-0.001), 'L0H4', np.float64(0.337)), (np.float64(-0.008), 'L0H0', np.float64(0.181))]

```