{
"id": "u18-amp-a0422-r1-q27b",
"title": "Unit 18 \u00b7 Fine sweep \u03b1=0.4221 \u00b7 qwen-27b \u00b7 RAND seed 1",
"unit": "18",
"created": "2026-08-07T22:57:50",
"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": 0.4221,
"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-a0422-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 water returns to the ground, flows back into bodies of water, and begins the journey again.\n\n**Note:** The prompt requires the description in two sentences. The first sentence covers evaporation and condensation, and the second sentence covers precipitation and return flow. To strictly adhere to the \"two sentences\" constraint, one would need to combine these processes into a single narrative flow, but the provided text above is a standard representation. If forced into exactly two sentences total, the description would look like this:\n\nThe water cycle is a continuous process where water evaporates from oceans and"
}
],
"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 water returns to the ground, flows back into bodies of water, and begins the journey again.\n\n**Note:** The prompt requires the description in two sentences. The first sentence covers evaporation and condensation, and the second sentence covers precipitation and return flow. To strictly adhere to the \"two sentences\" constraint, one would need to combine these processes into a single narrative flow, but the provided text above is a standard representation. If forced into exactly two sentences total, the description would look like this:\n\nThe water cycle is a continuous process where water evaporates from oceans and"
],
"tokens": [
"<|im_start|>",
"user",
"\n",
"Describe",
" the",
" water",
" cycle",
" in",
" two",
" sentences",
".",
"<|im_end|>",
"\n",
"<|im_start|>",
"assistant",
"\n",
"",
"\n\n",
"",
"\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",
" water",
" returns",
" to",
" the",
" ground",
",",
" flows",
" back",
" into",
" bodies",
" of",
" water",
",",
" and",
" begins",
" the",
" journey",
" again",
".",
"\n\n",
"**",
"Note",
":**",
" The",
" prompt",
" requires",
" the",
" description",
" in",
" two",
" sentences",
".",
" The",
" first",
" sentence",
" covers",
" ev",
"aporation",
" and",
" cond",
"ensation",
",",
" and",
" the",
" second",
" sentence",
" covers",
" precipitation",
" and",
" return",
" flow",
".",
" To",
" strictly",
" adhere",
" to",
" the",
" \"",
"two",
" sentences",
"\"",
" constraint",
",",
" one",
" would",
" need",
" to",
" combine",
" these",
" processes",
" into",
" a",
" single",
" narrative",
" flow",
",",
" but",
" the",
" provided",
" text",
" above",
" is",
" a",
" standard",
" representation",
".",
" If",
" forced",
" into",
" exactly",
" two",
" sentences",
" total",
",",
" the",
" description",
" would",
" look",
" like",
" this",
":",
"\n\n",
"The",
" water",
" cycle",
" is",
" a",
" continuous",
" process",
" where",
" water",
" evapor",
"ates",
" from",
" oceans",
" and",
"<|im_end|>",
"\n"
],
"readouts": [
{
"position": 170,
"token": "<|im_end|>",
"model_top": [
"\n",
"<|endoftext|>",
"\n\n",
" \n",
"\r\n",
"\n \n",
".",
" "
],
"layers": {
"0": [
" *@",
" ",
" **\u201c",
" \uff5c",
" ",
" $__",
"___",
"\r\n\r\n\r\n\r\n"
],
"16": [
"<|im_end|>",
" **\u3010",
"",
"\r\n\r\n\r\n\r\n",
"___",
"",
" ",
"",
" **\u3010",
"\r\n\r\n\r\n\r\n",
"",
"___",
"\n\n\n\n\n",
"\u548c"
],
"24": [
"<|im_end|>",
"\n\n",
"\r\n\r\n\r\n\r\n",
"\r\n\r\n",
"",
"<|endoftext|>",
"\n\n\n\n\n",
" \n\n"
],
"28": [
"<|im_end|>",
"\n\n",
"<|endoftext|>",
"\r\n\r\n\r\n\r\n",
"\r\n\r\n",
"\n\n\n\n\n",
" \n\n",
""
],
"32": [
"<|im_end|>",
"<|endoftext|>",
"",
"\r\n\r\n\r\n\r\n",
" **\u3010",
" \n\n",
"\n\n",
"\n\n\n\n\n"
],
"36": [
"<|im_end|>",
"<|endoftext|>",
"",
"\n\n",
"<|file_sep|>",
"\r\n\r\n\r\n\r\n",
"<|im_start|>",
" \n\n"
],
"40": [
"<|endoftext|>",
"<|im_end|>",
"",
" \n\n",
"<|file_sep|>",
"<|im_start|>",
"\r\n\r\n\r\n\r\n",
" \n\n"
],
"44": [
"<|endoftext|>",
"<|im_end|>",
"",
"<|file_sep|>",
"\r\n\r\n\r\n\r\n",
"\u3002",
" **\u3010",
" \n\n"
],
"48": [
"<|im_end|>",
")\uff0c",
"\u3002",
"\uff0c",
"\u5728",
"\u4e2d",
"\u548c",
")\u3002"
],
"52": [
"<|endoftext|>",
"<|im_end|>",
"<|im_start|>",
"",
"<|file_sep|>",
"<|fim_prefix|>",
"<|fim_middle|>",
";**"
],
"56": [
"<|im_end|>",
"<|im_start|>",
"<|endoftext|>",
"",
"<|fim_prefix|>",
"-placeholder",
")\">",
"(Output"
],
"58": [
"<|im_end|>",
"<|endoftext|>",
"<|im_start|>",
"",
"\u8fde",
"<|fim_prefix|>",
"(tk",
"\u7f51"
],
"60": [
"<|endoftext|>",
"<|im_start|>",
"<|im_end|>",
"NETWORK",
"\u7f51",
"",
"\u9274",
"assistant"
],
"62": [
"\n",
" \n",
"\n \n",
"<|im_end|>",
"<|im_start|>",
"\n \n",
" \n",
"\n \n"
]
}
}
],
"trajectories": [
{
"word": "anyways",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
991,
6401,
18287,
15538,
5711,
10032,
9302,
13591,
20917,
40843,
65116,
132825,
178892,
203950,
140157,
46971,
79901,
26795
]
},
{
"word": "alot",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
121,
4186,
11897,
5877,
1245,
4533,
2584,
3513,
5043,
14047,
20816,
12340,
54879,
78216,
17126,
20390,
24850,
9600
]
},
{
"word": "yummy",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
17400,
8537,
27025,
11034,
12691,
69383,
61202,
64212,
50157,
120907,
166203,
197469,
106940,
222601,
172661,
146235,
148531,
121785
]
},
{
"word": "kinda",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
4184,
4600,
12499,
3339,
1969,
16668,
8451,
8328,
10226,
18095,
11140,
14828,
113601,
182319,
130350,
119740,
125737,
59013
]
},
{
"word": "whilst",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
13854,
103635,
167694,
73513,
10414,
8305,
1454,
1991,
86869,
36346,
96049,
77085,
155888,
111824,
80782,
78055,
89114,
39496
]
},
{
"word": "luckily",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
10924,
37599,
44473,
25823,
26681,
15610,
8044,
6290,
10506,
17609,
16947,
19202,
23665,
17371,
29126,
68079,
132159,
101956
]
}
],
"emergence": {
"position": 170,
"top1": "\n",
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
217254,
230728,
174162,
225696,
209808,
167590,
181584,
174860,
206966,
49100,
245439,
247333,
93006,
90341,
215344,
146393,
44649,
1
]
},
"scan": [],
"vanilla": {
"trajectories": [
{
"word": "anyways",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
169045,
156352,
130664,
109429,
158747,
157596,
150996,
147780,
59899,
67226,
144587,
150019,
194482,
200910,
98957,
77506,
55946,
45226
]
},
{
"word": "alot",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
130551,
129593,
123062,
118058,
73416,
120351,
65478,
79562,
199398,
233510,
222540,
120685,
84199,
33435,
5570,
6906,
4437,
5195
]
},
{
"word": "yummy",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
192171,
202845,
190309,
143958,
196828,
192029,
185926,
139551,
101275,
207177,
206196,
156858,
190607,
202420,
172952,
190868,
152809,
183149
]
},
{
"word": "kinda",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
90213,
90891,
54806,
46481,
30399,
55381,
88177,
128673,
64720,
37783,
74841,
29260,
163350,
213378,
141866,
134484,
149210,
94161
]
},
{
"word": "whilst",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
122544,
134213,
171876,
199052,
187560,
82621,
126249,
108047,
181383,
172279,
194871,
177073,
143408,
139297,
141833,
95564,
40297,
25207
]
},
{
"word": "luckily",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
154064,
118259,
115736,
89257,
77234,
35032,
35533,
45930,
75666,
73614,
70422,
32045,
32881,
7248,
154752,
179833,
180470,
132504
]
}
],
"top1_agreement": {
"0": 0.0,
"4": 0.0,
"8": 0.0,
"12": 0.0,
"16": 0.0,
"20": 0.0,
"24": 0.0,
"28": 0.0,
"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": 0.4221,
"32": 0.4221,
"36": 0.4221,
"40": 0.4221
}
]
}