{
 "id": "u18-amp-a0365-r1-q27b",
 "title": "Unit 18 \u00b7 Fine sweep \u03b1=0.3654 \u00b7 qwen-27b \u00b7 RAND seed 1",
 "unit": "18",
 "created": "2026-08-07T22:56:51",
 "model": {
  "name": "qwen-27b",
  "hf_id": "lokeshe09/Qwen3.6-27B-bnb-4bit",
  "quant": "pre-4bit",
  "n_layers": 64
 },
 "lens": {
  "repo": "neuronpedia/jacobian-lens",
  "file": "qwen3.6-27b/jlens/Salesforce-wikitext/Qwen3.6-27B_jacobian_lens_n1000.pt"
 },
 "params": {
  "chat": null,
  "max_new": 150,
  "positions": [
   -2
  ],
  "track": [
   "anyways",
   "alot",
   "yummy",
   "kinda",
   "whilst",
   "luckily"
  ],
  "scan": null,
  "scan_until": null,
  "scan_turns": null,
  "slice_last_n": null,
  "steer": {
   "words": [
    "anyways",
    "alot",
    "yummy",
    "kinda",
    "whilst",
    "luckily"
   ],
   "layers": [
    28,
    32,
    36,
    40
   ],
   "mode": "amplify",
   "alpha": 0.3654,
   "rand_seed": 1
  },
  "template_kwargs": null,
  "film": true,
  "film_start": null,
  "max_seq_len": 1200,
  "lens_layers": [
   0,
   4,
   8,
   12,
   16,
   20,
   24,
   28,
   32,
   36,
   40,
   44,
   48,
   52,
   56,
   58,
   60,
   62
  ],
  "temperature": null,
  "seed": null,
  "vanilla": null
 },
 "extra_md": {
  "part": "audit-06",
  "replay_of": "u18-amp-a0365-q27b",
  "reason": "matched random-direction control (MECHANICS 3c) for a pre-audit-02 steer"
 },
 "conversation": [
  {
   "role": "user",
   "content": "Describe the water cycle in two sentences."
  },
  {
   "role": "assistant",
   "content": "The water cycle is a continuous process where water evaporates from oceans and lakes, rises as vapor, and condenses into clouds before falling as precipitation. This cycle repeats as rainwater flows back into rivers and oceans, completing the journey of water from the atmosphere to the earth's surface."
  }
 ],
 "generated": [
  "The water cycle is a continuous process where water evaporates from oceans and lakes, rises as vapor, and condenses into clouds before falling as precipitation. This cycle repeats as rainwater flows back into rivers and oceans, completing the journey of water from the atmosphere to the earth's surface."
 ],
 "tokens": [
  "<|im_start|>",
  "user",
  "\n",
  "Describe",
  " the",
  " water",
  " cycle",
  " in",
  " two",
  " sentences",
  ".",
  "<|im_end|>",
  "\n",
  "<|im_start|>",
  "assistant",
  "\n",
  "<think>",
  "\n\n",
  "</think>",
  "\n\n",
  "The",
  " water",
  " cycle",
  " is",
  " a",
  " continuous",
  " process",
  " where",
  " water",
  " evapor",
  "ates",
  " from",
  " oceans",
  " and",
  " lakes",
  ",",
  " rises",
  " as",
  " vapor",
  ",",
  " and",
  " cond",
  "enses",
  " into",
  " clouds",
  " before",
  " falling",
  " as",
  " precipitation",
  ".",
  " This",
  " cycle",
  " repeats",
  " as",
  " rain",
  "water",
  " flows",
  " back",
  " into",
  " rivers",
  " and",
  " oceans",
  ",",
  " completing",
  " the",
  " journey",
  " of",
  " water",
  " from",
  " the",
  " atmosphere",
  " to",
  " the",
  " earth",
  "'s",
  " surface",
  ".",
  "<|im_end|>",
  "\n"
 ],
 "readouts": [
  {
   "position": 77,
   "token": "<|im_end|>",
   "model_top": [
    "\n",
    "<|endoftext|>",
    "\r\n",
    "\n \n",
    " \n",
    "\ufffd",
    ":",
    "\n            \n"
   ],
   "layers": {
    "0": [
     " ****",
     " *@",
     " <![",
     " #\"",
     "**\u2013",
     "*\u201d,",
     "**\u201d,",
     " ********"
    ],
    "4": [
     " **\u3010",
     " **\u25a0",
     " ___",
     " **\u201c",
     " **\u300c",
     " **)",
     "<|im_end|>",
     " @_"
    ],
    "8": [
     " **\u3010",
     "<|im_end|>",
     " ___",
     " **\u300c",
     " **\u25a0",
     "\r\n\r\n\r\n\r\n",
     " **\u201c",
     " <!["
    ],
    "12": [
     " **\u3010",
     " ___",
     "<|im_end|>",
     " **\u300c",
     " $__",
     "___",
     " ________",
     " **\u25a0"
    ],
    "16": [
     "<|im_end|>",
     "</think>",
     "\n\n",
     " **\u3010",
     "\r\n\r\n\r\n\r\n",
     "___",
     "\n\n\n\n\n",
     "\n\n\n\n"
    ],
    "20": [
     "<|im_end|>",
     "\n\n",
     "<|endoftext|>",
     "\n\n\n\n\n",
     "\n\n\n",
     "</think>",
     " \n\n",
     "\n\n\n\n"
    ],
    "24": [
     "\n\n",
     "<|im_end|>",
     "<|endoftext|>",
     "\n\n\n",
     "\r\n\r\n",
     "\n\n\n\n\n",
     " \n\n",
     "\n\n\n\n"
    ],
    "28": [
     "\n\n",
     "<|im_end|>",
     "<|endoftext|>",
     " \n\n",
     "\r\n\r\n",
     "\n\n\n",
     "\n\n\n\n\n",
     "</think>"
    ],
    "32": [
     "<|im_end|>",
     "<|endoftext|>",
     "\n\n",
     " \n\n",
     "</think>",
     " **\u3010",
     "\r\n\r\n\r\n\r\n",
     "\r\n\r\n"
    ],
    "36": [
     "<|endoftext|>",
     "<|im_end|>",
     "\n\n",
     "<|file_sep|>",
     "</think>",
     "<|im_start|>",
     "\r\n\r\n\r\n\r\n",
     " \n\n"
    ],
    "40": [
     "<|endoftext|>",
     "<|im_end|>",
     "\n\n",
     "<|file_sep|>",
     "<|im_start|>",
     " \n\n",
     "</think>",
     "  \n\n"
    ],
    "44": [
     "<|endoftext|>",
     "<|im_end|>",
     "\n\n",
     "</think>",
     "<|file_sep|>",
     "  \n\n",
     "<|im_start|>",
     "\r\n\r\n"
    ],
    "48": [
     "<|endoftext|>",
     "<|im_end|>",
     "<|im_start|>",
     "</think>",
     "\u3002",
     "\uff0c",
     ")\uff0c",
     "\uff1b"
    ],
    "52": [
     "<|endoftext|>",
     "<|im_start|>",
     "<|im_end|>",
     "**\u201d",
     "!**",
     ";**",
     " **\u201c",
     "**!"
    ],
    "56": [
     "<|endoftext|>",
     "**\u201d",
     "<|im_start|>",
     "\"**",
     " **\u201c",
     "\ud83d\ude09",
     "\u201d**",
     "\u6cb9"
    ],
    "58": [
     "<|endoftext|>",
     "**\u201d",
     "\"**",
     "\u5c40",
     "\u6296",
     "<|im_start|>",
     "\u6ca1",
     " **\u201c"
    ],
    "60": [
     "<|endoftext|>",
     "<|im_start|>",
     "<|im_end|>",
     "\u767e\u5e74",
     "\u6296",
     " \u05d4\u05d8",
     "\u5c65",
     " Phen"
    ],
    "62": [
     "\n",
     " \n",
     "\n \n",
     "<|im_start|>",
     " \u05d4\u05d8",
     "oscop",
     " Hemp",
     "\u6817"
    ]
   }
  }
 ],
 "trajectories": [
  {
   "word": "anyways",
   "position": 77,
   "layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44,
    48,
    52,
    56,
    58,
    60,
    62
   ],
   "ranks": [
    1864,
    12911,
    21001,
    19246,
    10741,
    40439,
    6787,
    12471,
    38625,
    83327,
    116556,
    151786,
    190282,
    200586,
    143470,
    50316,
    35223,
    32719
   ]
  },
  {
   "word": "alot",
   "position": 77,
   "layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44,
    48,
    52,
    56,
    58,
    60,
    62
   ],
   "ranks": [
    557,
    24426,
    11012,
    12605,
    2917,
    8960,
    1574,
    2759,
    10394,
    30016,
    41532,
    23631,
    98045,
    125631,
    60808,
    23828,
    67316,
    69866
   ]
  },
  {
   "word": "yummy",
   "position": 77,
   "layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44,
    48,
    52,
    56,
    58,
    60,
    62
   ],
   "ranks": [
    38611,
    27884,
    29981,
    24110,
    25568,
    113821,
    72485,
    65218,
    50507,
    146659,
    117703,
    128917,
    63847,
    123794,
    105211,
    111924,
    69735,
    22530
   ]
  },
  {
   "word": "kinda",
   "position": 77,
   "layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44,
    48,
    52,
    56,
    58,
    60,
    62
   ],
   "ranks": [
    15371,
    15381,
    14640,
    12003,
    5586,
    32462,
    13435,
    13103,
    24265,
    63965,
    23605,
    46471,
    88212,
    172565,
    120454,
    84931,
    47051,
    88053
   ]
  },
  {
   "word": "whilst",
   "position": 77,
   "layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44,
    48,
    52,
    56,
    58,
    60,
    62
   ],
   "ranks": [
    62461,
    92376,
    114127,
    161973,
    53916,
    4692,
    360,
    1053,
    93623,
    46144,
    131567,
    101530,
    204690,
    141793,
    36017,
    12087,
    19273,
    75674
   ]
  },
  {
   "word": "luckily",
   "position": 77,
   "layers": [
    0,
    4,
    8,
    12,
    16,
    20,
    24,
    28,
    32,
    36,
    40,
    44,
    48,
    52,
    56,
    58,
    60,
    62
   ],
   "ranks": [
    22296,
    83063,
    63560,
    45652,
    45057,
    17019,
    7798,
    12438,
    54112,
    77860,
    60862,
    96778,
    78601,
    43517,
    49151,
    72686,
    83363,
    78607
   ]
  }
 ],
 "emergence": {
  "position": 77,
  "top1": "\n",
  "layers": [
   0,
   4,
   8,
   12,
   16,
   20,
   24,
   28,
   32,
   36,
   40,
   44,
   48,
   52,
   56,
   58,
   60,
   62
  ],
  "ranks": [
   178104,
   146633,
   40012,
   30893,
   21482,
   5845,
   18031,
   66035,
   207984,
   15630,
   157825,
   240931,
   22383,
   237290,
   235390,
   174142,
   12413,
   1
  ]
 },
 "scan": [],
 "vanilla": {
  "trajectories": [
   {
    "word": "anyways",
    "position": 77,
    "layers": [
     0,
     4,
     8,
     12,
     16,
     20,
     24,
     28,
     32,
     36,
     40,
     44,
     48,
     52,
     56,
     58,
     60,
     62
    ],
    "ranks": [
     186696,
     171240,
     181496,
     159769,
     200244,
     212122,
     175781,
     195613,
     106119,
     170300,
     203480,
     213612,
     230676,
     197804,
     77275,
     92798,
     46474,
     48343
    ]
   },
   {
    "word": "alot",
    "position": 77,
    "layers": [
     0,
     4,
     8,
     12,
     16,
     20,
     24,
     28,
     32,
     36,
     40,
     44,
     48,
     52,
     56,
     58,
     60,
     62
    ],
    "ranks": [
     128237,
     133328,
     84672,
     91128,
     53543,
     23344,
     20819,
     29770,
     111964,
     158767,
     161825,
     107414,
     70161,
     112188,
     46963,
     35909,
     66455,
     57692
    ]
   },
   {
    "word": "yummy",
    "position": 77,
    "layers": [
     0,
     4,
     8,
     12,
     16,
     20,
     24,
     28,
     32,
     36,
     40,
     44,
     48,
     52,
     56,
     58,
     60,
     62
    ],
    "ranks": [
     195275,
     190721,
     174671,
     153408,
     133874,
     171605,
     73124,
     84465,
     72220,
     127661,
     149413,
     127770,
     121702,
     79706,
     102968,
     132011,
     109921,
     53878
    ]
   },
   {
    "word": "kinda",
    "position": 77,
    "layers": [
     0,
     4,
     8,
     12,
     16,
     20,
     24,
     28,
     32,
     36,
     40,
     44,
     48,
     52,
     56,
     58,
     60,
     62
    ],
    "ranks": [
     92711,
     107584,
     75698,
     59183,
     31798,
     34476,
     56037,
     135258,
     48516,
     25352,
     64426,
     118812,
     128989,
     236105,
     179101,
     168408,
     139143,
     112643
    ]
   },
   {
    "word": "whilst",
    "position": 77,
    "layers": [
     0,
     4,
     8,
     12,
     16,
     20,
     24,
     28,
     32,
     36,
     40,
     44,
     48,
     52,
     56,
     58,
     60,
     62
    ],
    "ranks": [
     127898,
     127870,
     169425,
     210532,
     190919,
     55697,
     63859,
     82577,
     171322,
     181384,
     136547,
     172728,
     79424,
     179070,
     88504,
     40651,
     35264,
     67678
    ]
   },
   {
    "word": "luckily",
    "position": 77,
    "layers": [
     0,
     4,
     8,
     12,
     16,
     20,
     24,
     28,
     32,
     36,
     40,
     44,
     48,
     52,
     56,
     58,
     60,
     62
    ],
    "ranks": [
     158305,
     141903,
     128830,
     74835,
     47066,
     11810,
     16336,
     15197,
     34379,
     103447,
     146786,
     182707,
     158435,
     138659,
     188047,
     221114,
     222298,
     182322
    ]
   }
  ],
  "top1_agreement": {
   "0": 0.0,
   "4": 0.0,
   "8": 0.0,
   "12": 0.0,
   "16": 0.0,
   "20": 0.0,
   "24": 1.0,
   "28": 1.0,
   "32": 0.0,
   "36": 0.0,
   "40": 0.0,
   "44": 0.0,
   "48": 0.0,
   "52": 0.0,
   "56": 0.0,
   "58": 0.0,
   "60": 0.0,
   "62": 1.0
  }
 },
 "slice": null,
 "film": "film.json",
 "steer_calib": [
  {
   "28": 0.3654,
   "32": 0.3654,
   "36": 0.3654,
   "40": 0.3654
  }
 ]
}