{
"id": "u18-amp-a0480-r1-q27b",
"title": "Unit 18 \u00b7 Fine sweep \u03b1=0.48 \u00b7 qwen-27b \u00b7 RAND seed 1",
"unit": "18",
"created": "2026-08-07T22:59:11",
"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": 0.48,
"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-a0480-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 rivers, rises into the atmosphere as vapor, and then condenses into clouds. When precipitation occurs, the water falls back to the earth, completing the cycle as it returns to the oceans and rivers.\n\n**Note:** The prompt asks to describe the water cycle in two sentences. The provided text is a single sentence that describes the entire cycle. To fulfill the constraint, one could combine the two sentences into one by merging the clauses.\n\n**Revised Version (Two Sentences Constraint):**\nThe water cycle is a continuous process where water evaporates from oceans and rivers, rises into the atmosphere as vapor, and then condenses into clouds. When precipitation occurs, the water falls"
}
],
"generated": [
"The water cycle is a continuous process where water evaporates from oceans and rivers, rises into the atmosphere as vapor, and then condenses into clouds. When precipitation occurs, the water falls back to the earth, completing the cycle as it returns to the oceans and rivers.\n\n**Note:** The prompt asks to describe the water cycle in two sentences. The provided text is a single sentence that describes the entire cycle. To fulfill the constraint, one could combine the two sentences into one by merging the clauses.\n\n**Revised Version (Two Sentences Constraint):**\nThe water cycle is a continuous process where water evaporates from oceans and rivers, rises into the atmosphere as vapor, and then condenses into clouds. When precipitation occurs, the water falls"
],
"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",
" rivers",
",",
" rises",
" into",
" the",
" atmosphere",
" as",
" vapor",
",",
" and",
" then",
" cond",
"enses",
" into",
" clouds",
".",
" When",
" precipitation",
" occurs",
",",
" the",
" water",
" falls",
" back",
" to",
" the",
" earth",
",",
" completing",
" the",
" cycle",
" as",
" it",
" returns",
" to",
" the",
" oceans",
" and",
" rivers",
".",
"\n\n",
"**",
"Note",
":**",
" The",
" prompt",
" asks",
" to",
" describe",
" the",
" water",
" cycle",
" in",
" two",
" sentences",
".",
" The",
" provided",
" text",
" is",
" a",
" single",
" sentence",
" that",
" describes",
" the",
" entire",
" cycle",
".",
" To",
" fulfill",
" the",
" constraint",
",",
" one",
" could",
" combine",
" the",
" two",
" sentences",
" into",
" one",
" by",
" merging",
" the",
" clauses",
".",
"\n\n",
"**",
"Re",
"vised",
" Version",
" (",
"Two",
" Sent",
"ences",
" Constraint",
"):**",
"\n",
"The",
" water",
" cycle",
" is",
" a",
" continuous",
" process",
" where",
" water",
" evapor",
"ates",
" from",
" oceans",
" and",
" rivers",
",",
" rises",
" into",
" the",
" atmosphere",
" as",
" vapor",
",",
" and",
" then",
" cond",
"enses",
" into",
" clouds",
".",
" When",
" precipitation",
" occurs",
",",
" the",
" water",
" falls",
"<|im_end|>",
"\n"
],
"readouts": [
{
"position": 170,
"token": "<|im_end|>",
"model_top": [
"\n",
"<|endoftext|>",
"\n\n",
" \n",
" ",
".",
",",
"\n \n"
],
"layers": {
"0": [
"**\u2013",
"*\u201d,",
" ****",
"**\u201d,",
" *@",
" ",
" **\u300c",
" ___",
" **\u25a0",
" ",
" **\u25a0",
" $__",
"___",
" __("
],
"16": [
"<|im_end|>",
" **\u3010",
"",
"___",
"\r\n\r\n\r\n\r\n",
" ___",
" "
],
"20": [
"<|im_end|>",
" **\u3010",
"",
"\r\n\r\n\r\n\r\n",
"___",
"",
"\n\n",
"\n\n\n\n\n"
],
"24": [
"<|im_end|>",
"\n\n",
"\r\n\r\n\r\n\r\n",
"\r\n\r\n",
" \n\n",
"",
"___",
" \r\n\r\n"
],
"28": [
"<|im_end|>",
"<|endoftext|>",
"\n\n",
" \n\n",
"\r\n\r\n",
"\r\n\r\n\r\n\r\n",
" \r\n\r\n",
"\n\n\n\n\n"
],
"32": [
"<|im_end|>",
"<|endoftext|>",
" \n\n",
"",
"\r\n\r\n\r\n\r\n",
"\n\n",
" **\u3010",
" \n\n"
],
"36": [
"<|im_end|>",
"<|endoftext|>",
"\n\n",
"",
" \n\n",
"<|file_sep|>",
"<|im_start|>",
"\r\n\r\n\r\n\r\n"
],
"40": [
"<|endoftext|>",
"<|im_end|>",
" \n\n",
" \n\n",
"",
"\n\n",
"<|im_start|>",
"\r\n\r\n\r\n\r\n"
],
"44": [
"<|im_end|>",
"<|endoftext|>",
"",
" \n\n",
" **\u3010",
" \n\n",
"\r\n\r\n\r\n\r\n",
"\n\n"
],
"48": [
"<|im_end|>",
"\u3002",
"\uff0c",
")\uff0c",
"\u5728",
"\u548c",
"<|endoftext|>",
")\u3002"
],
"52": [
"<|endoftext|>",
"<|im_end|>",
"<|im_start|>",
"",
"<|file_sep|>",
" \n\n",
"<|fim_prefix|>",
"\n\n"
],
"56": [
"<|im_end|>",
"<|im_start|>",
"",
"<|endoftext|>",
"FAILURE",
"<|fim_prefix|>",
"lobs",
"<|quad_start|>"
],
"58": [
"<|im_end|>",
"<|endoftext|>",
"<|im_start|>",
"",
"\u9000",
"\u67cf",
"\u8131",
"(Output"
],
"60": [
"<|endoftext|>",
"<|im_start|>",
"assistant",
"<|im_end|>",
"\u52a9\u7406",
"\u52a9",
" assistant",
""
],
"62": [
"\n",
" \n",
"\n \n",
"\n \n",
"<|im_start|>",
"<|im_end|>",
" \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": [
1256,
6580,
19713,
19862,
9676,
27135,
18751,
30930,
34281,
77272,
103623,
148732,
224999,
194162,
167844,
37080,
63043,
32515
]
},
{
"word": "alot",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
234,
7077,
11633,
8405,
1306,
8298,
3479,
4438,
6174,
11742,
13439,
7951,
40461,
64293,
35226,
45300,
42506,
12111
]
},
{
"word": "yummy",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
24723,
10807,
27153,
12395,
12366,
66223,
58556,
63488,
46294,
94621,
117063,
179823,
113729,
215527,
138028,
107242,
115307,
133570
]
},
{
"word": "kinda",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
6648,
6846,
10979,
4990,
2906,
26800,
12641,
10431,
10842,
14577,
6008,
8631,
86134,
158715,
113966,
73244,
77237,
27831
]
},
{
"word": "whilst",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
21134,
131079,
206217,
157511,
23267,
27988,
2714,
2750,
91481,
45692,
116947,
79192,
108140,
55711,
44322,
29287,
58903,
62331
]
},
{
"word": "luckily",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
14673,
36554,
39051,
31595,
25864,
20250,
10875,
7722,
13891,
18905,
17665,
29545,
15996,
22618,
22811,
43016,
61281,
39520
]
}
],
"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": [
218062,
231845,
91998,
206496,
190203,
191257,
199228,
190322,
235464,
72072,
238506,
247669,
71696,
12867,
163196,
51387,
1973,
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": [
174484,
152664,
149777,
143218,
148734,
157498,
141476,
181910,
78279,
104718,
174542,
141312,
217703,
205002,
152211,
84967,
65933,
58685
]
},
{
"word": "alot",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
132428,
137043,
102399,
105797,
80602,
136151,
68017,
83640,
195540,
237820,
230908,
118084,
90823,
24808,
9634,
21779,
14159,
7577
]
},
{
"word": "yummy",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
191456,
198156,
189643,
150573,
183645,
166767,
184929,
148526,
93448,
211519,
206805,
195968,
174142,
203838,
106962,
151296,
124712,
183955
]
},
{
"word": "kinda",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
87458,
104085,
62797,
62919,
61481,
84573,
117362,
172976,
68072,
53676,
81640,
21391,
132459,
225140,
163172,
117704,
159871,
53202
]
},
{
"word": "whilst",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
124605,
119113,
168440,
206944,
203583,
114173,
143172,
136724,
185128,
166364,
179196,
168006,
138982,
84298,
80407,
53244,
46387,
69375
]
},
{
"word": "luckily",
"position": 170,
"layers": [
0,
4,
8,
12,
16,
20,
24,
28,
32,
36,
40,
44,
48,
52,
56,
58,
60,
62
],
"ranks": [
157341,
157108,
142045,
119304,
111684,
37137,
52004,
58516,
89858,
77026,
69598,
26954,
23473,
10710,
111995,
133100,
92520,
41358
]
}
],
"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": 0.48,
"32": 0.48,
"36": 0.48,
"40": 0.48
}
]
}