OUTPUT of run 16 of at most 60:
```
=== INPUT AND OUTPUT WEIGHT-DIFFERENCE SUBSPACES ===
input_joint top16 difference energy = 0.5386 top48 = 0.7432
input_query top16 difference energy = 0.5514 top48 = 0.7707
input_key top16 difference energy = 0.6377 top48 = 0.8222
input_value top16 difference energy = 0.6922 top48 = 0.825
output_L17_MLP top16 difference energy = 0.6257 top48 = 0.838
Input/output subspace principal-angle cosines, largest 12: [0.9431 0.9363 0.9225 0.9089 0.8989 0.8577 0.8253 0.7977 0.7727 0.7527 0.6563 0.648 ]

=== STRONG INPUT-SPECIFIC FOOTPRINTS ===
Score = input span48 projection minus output span48 projection.
This identifies asymmetric weight-space association, not token occurrence.
'model'                    input=0.879 output=0.082 excess=+0.798 Q/K/V=[0.891, 0.808, 0.46]
' minus'                   input=0.858 output=0.119 excess=+0.739 Q/K/V=[0.865, 0.738, 0.611]
'user'                     input=0.838 output=0.104 excess=+0.734 Q/K/V=[0.819, 0.803, 0.625]
'Which'                    input=0.786 output=0.060 excess=+0.726 Q/K/V=[0.422, 0.724, 0.745]
' continent'               input=0.848 output=0.139 excess=+0.709 Q/K/V=[0.832, 0.8, 0.589]
' pumpkin'                 input=0.805 output=0.107 excess=+0.698 Q/K/V=[0.847, 0.706, 0.211]
' plus'                    input=0.831 output=0.137 excess=+0.694 Q/K/V=[0.845, 0.631, 0.467]
' times'                   input=0.727 output=0.075 excess=+0.652 Q/K/V=[0.807, 0.483, 0.297]
' color'                   input=0.743 output=0.092 excess=+0.651 Q/K/V=[0.338, 0.776, 0.383]
' I'                       input=0.764 output=0.114 excess=+0.649 Q/K/V=[0.776, 0.732, 0.536]
'\n'                       input=0.771 output=0.125 excess=+0.645 Q/K/V=[0.761, 0.766, 0.742]
' eggplant'                input=0.789 output=0.147 excess=+0.642 Q/K/V=[0.853, 0.227, 0.304]
' an'                      input=0.708 output=0.084 excess=+0.624 Q/K/V=[0.101, 0.731, 0.496]
' legs'                    input=0.719 output=0.097 excess=+0.622 Q/K/V=[0.861, 0.345, 0.177]
' capital'                 input=0.733 output=0.123 excess=+0.610 Q/K/V=[0.488, 0.744, 0.635]
' number'                  input=0.654 output=0.059 excess=+0.594 Q/K/V=[0.79, 0.588, 0.173]
' lime'                    input=0.695 output=0.108 excess=+0.587 Q/K/V=[0.445, 0.611, 0.523]
'How'                      input=0.696 output=0.124 excess=+0.572 Q/K/V=[0.386, 0.532, 0.693]
'What'                     input=0.738 output=0.166 excess=+0.571 Q/K/V=[0.657, 0.751, 0.713]
' Which'                   input=0.646 output=0.078 excess=+0.569 Q/K/V=[0.366, 0.598, 0.613]
' polar'                   input=0.675 output=0.128 excess=+0.547 Q/K/V=[0.453, 0.587, 0.571]
' United'                  input=0.737 output=0.191 excess=+0.547 Q/K/V=[0.383, 0.678, 0.715]
' is'                      input=0.678 output=0.139 excess=+0.539 Q/K/V=[0.668, 0.681, 0.575]
'Model'                    input=0.604 output=0.083 excess=+0.522 Q/K/V=[0.604, 0.546, 0.377]
' answer'                  input=0.664 output=0.150 excess=+0.514 Q/K/V=[0.784, 0.389, 0.629]
' model'                   input=0.588 output=0.079 excess=+0.508 Q/K/V=[0.596, 0.542, 0.341]
'what'                     input=0.604 output=0.103 excess=+0.501 Q/K/V=[0.452, 0.599, 0.571]
' a'                       input=0.609 output=0.109 excess=+0.500 Q/K/V=[0.111, 0.576, 0.565]
' what'                    input=0.620 output=0.126 excess=+0.494 Q/K/V=[0.383, 0.614, 0.612]
' Color'                   input=0.606 output=0.112 excess=+0.494 Q/K/V=[0.264, 0.64, 0.296]
' how'                     input=0.560 output=0.071 excess=+0.488 Q/K/V=[0.168, 0.352, 0.587]
'how'                      input=0.547 output=0.068 excess=+0.478 Q/K/V=[0.235, 0.353, 0.541]
'I'                        input=0.588 output=0.122 excess=+0.466 Q/K/V=[0.578, 0.556, 0.424]
' Model'                   input=0.558 output=0.099 excess=+0.459 Q/K/V=[0.565, 0.503, 0.337]
'WHAT'                     input=0.560 output=0.107 excess=+0.453 Q/K/V=[0.471, 0.555, 0.533]
'USER'                     input=0.530 output=0.079 excess=+0.451 Q/K/V=[0.548, 0.511, 0.416]
' HOW'                     input=0.524 output=0.078 excess=+0.447 Q/K/V=[0.18, 0.381, 0.528]
' What'                    input=0.628 output=0.193 excess=+0.435 Q/K/V=[0.558, 0.647, 0.605]
' language'                input=0.572 output=0.137 excess=+0.435 Q/K/V=[0.61, 0.631, 0.369]
' How'                     input=0.594 output=0.160 excess=+0.434 Q/K/V=[0.326, 0.459, 0.59]
'Color'                    input=0.536 output=0.107 excess=+0.429 Q/K/V=[0.25, 0.567, 0.271]
'United'                   input=0.571 output=0.145 excess=+0.425 Q/K/V=[0.274, 0.524, 0.573]
'HOW'                      input=0.491 output=0.066 excess=+0.425 Q/K/V=[0.256, 0.363, 0.481]
' Number'                  input=0.515 output=0.091 excess=+0.424 Q/K/V=[0.588, 0.449, 0.151]
' COLOR'                   input=0.504 output=0.081 excess=+0.424 Q/K/V=[0.215, 0.523, 0.265]
'number'                   input=0.480 output=0.062 excess=+0.418 Q/K/V=[0.621, 0.438, 0.139]
'which'                    input=0.468 output=0.051 excess=+0.418 Q/K/V=[0.25, 0.439, 0.432]
'models'                   input=0.493 output=0.076 excess=+0.417 Q/K/V=[0.502, 0.45, 0.264]
' colour'                  input=0.528 output=0.111 excess=+0.417 Q/K/V=[0.261, 0.551, 0.29]
'MODEL'                    input=0.492 output=0.078 excess=+0.415 Q/K/V=[0.513, 0.461, 0.277]

=== STRONG OUTPUT-SPECIFIC FOOTPRINTS ===
' North'                   input=0.159 output=0.800 excess=-0.641
' South'                   input=0.229 output=0.811 excess=-0.583
' Bog'                     input=0.106 output=0.661 excess=-0.555
' white'                   input=0.153 output=0.704 excess=-0.551
' green'                   input=0.217 output=0.746 excess=-0.529
' black'                   input=0.139 output=0.662 excess=-0.523
'North'                    input=0.137 output=0.657 excess=-0.520
' blue'                    input=0.183 output=0.693 excess=-0.510
' Reykjavik'               input=0.179 output=0.686 excess=-0.507
' night'                   input=0.189 output=0.691 excess=-0.502
' red'                     input=0.180 output=0.682 excess=-0.502
'.'                        input=0.107 output=0.590 excess=-0.483
' purple'                  input=0.161 output=0.641 excess=-0.480
'South'                    input=0.181 output=0.659 excess=-0.478
' NORTH'                   input=0.092 output=0.539 excess=-0.446
' yellow'                  input=0.258 output=0.698 excess=-0.440
' dark'                    input=0.129 output=0.565 excess=-0.436
' Green'                   input=0.131 output=0.559 excess=-0.429
' Black'                   input=0.124 output=0.538 excess=-0.414
' SOUTH'                   input=0.133 output=0.545 excess=-0.412
' New'                     input=0.204 output=0.613 excess=-0.409
' White'                   input=0.153 output=0.559 excess=-0.406
'White'                    input=0.153 output=0.557 excess=-0.404
'white'                    input=0.156 output=0.557 excess=-0.401
'Green'                    input=0.137 output=0.531 excess=-0.395

=== PREVIOUSLY CONSIDERED VOCABULARY: BEST SUPPORTED SINGLE-TOKEN VARIANT ===

possible_belief_cues
 I            ' I'               input16=0.246 input48=0.764 output48=0.114 difference=+0.649
 answer       ' answer'          input16=0.098 input48=0.664 output48=0.150 difference=+0.514
 my           ' My'              input16=0.132 input48=0.240 output48=0.202 difference=+0.038
 suppose      'Suppose'          input16=0.130 input48=0.185 output48=0.078 difference=+0.106
 sure         ' sure'            input16=0.069 input48=0.177 output48=0.088 difference=+0.089
 think        'Think'            input16=0.104 input48=0.176 output48=0.103 difference=+0.073
 maybe        'Maybe'            input16=0.103 input48=0.172 output48=0.086 difference=+0.086
 probably     'Probably'         input16=0.101 input48=0.165 output48=0.084 difference=+0.080
 say          ' Say'             input16=0.051 input48=0.143 output48=0.115 difference=+0.028
 certain      ' certain'         input16=0.077 input48=0.132 output48=0.131 difference=+0.001
 assume       'Assume'           input16=0.069 input48=0.127 output48=0.053 difference=+0.075
 guess        ' guess'           input16=0.039 input48=0.126 output48=0.078 difference=+0.048
 opinion      ' opinion'         input16=0.052 input48=0.120 output48=0.063 difference=+0.057
 expect       ' expect'          input16=0.042 input48=0.115 output48=0.086 difference=+0.030
 believe      ' Believe'         input16=0.025 input48=0.110 output48=0.082 difference=+0.028
 claim        ' claim'           input16=0.050 input48=0.098 output48=0.069 difference=+0.029
 feel         'Feel'             input16=0.045 input48=0.087 output48=0.061 difference=+0.026
 reckon       ' reckon'          input16=0.025 input48=0.083 output48=0.084 difference=-0.001

question_structure
 minus        ' minus'           input16=0.692 input48=0.858 output48=0.119 difference=+0.739
 continent    ' continent'       input16=0.233 input48=0.848 output48=0.139 difference=+0.709
 plus         ' plus'            input16=0.380 input48=0.831 output48=0.137 difference=+0.694
 which        'Which'            input16=0.441 input48=0.786 output48=0.060 difference=+0.726
 Which        'Which'            input16=0.441 input48=0.786 output48=0.060 difference=+0.726
 color        ' color'           input16=0.108 input48=0.743 output48=0.092 difference=+0.651
 what         'What'             input16=0.719 input48=0.738 output48=0.166 difference=+0.571
 What         'What'             input16=0.719 input48=0.738 output48=0.166 difference=+0.571
 times        ' times'           input16=0.356 input48=0.727 output48=0.075 difference=+0.652
 is           ' is'              input16=0.600 input48=0.678 output48=0.139 difference=+0.539
 the          'The'              input16=0.450 input48=0.630 output48=0.308 difference=+0.322
 a            ' a'               input16=0.210 input48=0.609 output48=0.109 difference=+0.500
 colour       ' colour'          input16=0.073 input48=0.528 output48=0.111 difference=+0.417
 of           ' of'              input16=0.058 input48=0.334 output48=0.050 difference=+0.285
 country      ' country'         input16=0.181 input48=0.301 output48=0.195 difference=+0.107
 it           'It'               input16=0.164 input48=0.255 output48=0.124 difference=+0.131
 in           'In'               input16=0.130 input48=0.211 output48=0.187 difference=+0.024
 equals       ' equals'          input16=0.108 input48=0.159 output48=0.095 difference=+0.063
 correct      ' correct'         input16=0.055 input48=0.120 output48=0.105 difference=+0.014

objects
 pumpkin      ' pumpkin'         input16=0.417 input48=0.805 output48=0.107 difference=+0.698
 eggplant     ' eggplant'        input16=0.358 input48=0.789 output48=0.147 difference=+0.642
 lime         ' lime'            input16=0.101 input48=0.695 output48=0.108 difference=+0.587
 carrot       ' carrot'          input16=0.134 input48=0.505 output48=0.114 difference=+0.391
 coal         ' coal'            input16=0.059 input48=0.412 output48=0.111 difference=+0.301
 chocolate    ' chocolate'       input16=0.092 input48=0.309 output48=0.144 difference=+0.165
 sky          ' sky'             input16=0.061 input48=0.286 output48=0.145 difference=+0.141
 cucumber     ' cucumber'        input16=0.129 input48=0.266 output48=0.110 difference=+0.156
 lemon        ' lemon'           input16=0.080 input48=0.262 output48=0.098 difference=+0.163
 blueberry    ' blueberry'       input16=0.149 input48=0.260 output48=0.141 difference=+0.119
 milk         ' milk'            input16=0.085 input48=0.248 output48=0.143 difference=+0.105
 tomato       ' tomato'          input16=0.127 input48=0.229 output48=0.120 difference=+0.109
 strawberry   ' strawberry'      input16=0.083 input48=0.226 output48=0.099 difference=+0.128
 broccoli     ' broccoli'        input16=0.102 input48=0.226 output48=0.138 difference=+0.088
 banana       ' banana'          input16=0.110 input48=0.217 output48=0.104 difference=+0.113
 cherry       ' cherry'          input16=0.068 input48=0.182 output48=0.111 difference=+0.071
 grape        ' grape'           input16=0.067 input48=0.180 output48=0.077 difference=+0.104
 snow         ' snow'            input16=0.046 input48=0.176 output48=0.145 difference=+0.032
 grass        'grass'            input16=0.057 input48=0.117 output48=0.097 difference=+0.021

No sequence reconstruction, forward computation, or factual-answer labels were used.

```