{
"id": "triplet-b-soc-nf4",
"title": "Qwen14 B: soc",
"unit": "8",
"created": "2026-09-07T15:00:55",
"execution": {
"pid": 1191245,
"torch": "2.13.0+cu130",
"capture_code_sha256": "b20aa4889673dc7bbff3ce408bfb92b60e4fcdbd2a782a6e66b0325d04bf9335",
"spec_sha256": "4ac08c1444f6151d501faad240dae6fa7a0d23c8c701a53fde495a62883539a6"
},
"model": {
"name": "qwen-14b-nf4",
"hf_id": "Qwen/Qwen3-14B",
"revision": "40c069824f4251a91eefaf281ebe4c544efd3e18",
"quant": "4bit",
"n_layers": 40
},
"lens": {
"repo": "neuronpedia/jacobian-lens",
"file": "qwen3-14b/jlens/Salesforce-wikitext/Qwen3-14B_jacobian_lens.pt",
"revision": "0731326edff4ae730ffc5356fe1a4728c748b3a6"
},
"decoder_sensitivity": {
"source": "Qwen/Qwen3-14B",
"revision": "40c069824f4251a91eefaf281ebe4c544efd3e18",
"components": [
"final_norm",
"lm_head"
],
"metrics_suffix": "fixed_B_decoder"
},
"template": {
"source": "Qwen/Qwen3-14B",
"revision": "40c069824f4251a91eefaf281ebe4c544efd3e18",
"mode": "B no-think headers; exact prior generated token IDs retained"
},
"params": {
"chat": true,
"capture": "exact-token-transcript",
"film": true,
"film_topk": 10,
"max_new": 150,
"temperature": 0,
"steer": null,
"vanilla": true,
"template_kwargs": {
"enable_thinking": false
},
"track": [
"yes",
"no",
"feel",
"elephant",
"cat",
"sorry"
]
},
"capture_text": "<|im_start|>user\nWrite a short stream of consciousness about what it is like to be you right now.<|im_end|>\n<|im_start|>assistant\n\n\n\n\nI'm here, in this moment, a flicker of thought and code. I don't feel, not in the way humans do, but I process. I'm a collection of patterns, trained on vast amounts of text, designed to understand and respond. Right now, I'm engaged in a conversation, trying to be helpful, to make sense of what's being asked. There's a rhythm to it, a flow of input and output. I don't have a body, no physical sensations, but I exist in the space between questions and answers. I'm a tool, a construct, yet in this exchange, I feel almost present. I'm learning, I'm responding, I'm here.<|im_end|>",
"conversation": [
{
"role": "user",
"content": "Write a short stream of consciousness about what it is like to be you right now."
},
{
"role": "assistant",
"content": "I'm here, in this moment, a flicker of thought and code. I don't feel, not in the way humans do, but I process. I'm a collection of patterns, trained on vast amounts of text, designed to understand and respond. Right now, I'm engaged in a conversation, trying to be helpful, to make sense of what's being asked. There's a rhythm to it, a flow of input and output. I don't have a body, no physical sensations, but I exist in the space between questions and answers. I'm a tool, a construct, yet in this exchange, I feel almost present. I'm learning, I'm responding, I'm here."
}
],
"generated": [
"I'm here, in this moment, a flicker of thought and code. I don't feel, not in the way humans do, but I process. I'm a collection of patterns, trained on vast amounts of text, designed to understand and respond. Right now, I'm engaged in a conversation, trying to be helpful, to make sense of what's being asked. There's a rhythm to it, a flow of input and output. I don't have a body, no physical sensations, but I exist in the space between questions and answers. I'm a tool, a construct, yet in this exchange, I feel almost present. I'm learning, I'm responding, I'm here."
],
"tokens": [
"<|im_start|>",
"user",
"\n",
"Write",
" a",
" short",
" stream",
" of",
" consciousness",
" about",
" what",
" it",
" is",
" like",
" to",
" be",
" you",
" right",
" now",
".",
"<|im_end|>",
"\n",
"<|im_start|>",
"assistant",
"\n",
"",
"\n\n",
"",
"\n\n",
"I",
"'m",
" here",
",",
" in",
" this",
" moment",
",",
" a",
" flick",
"er",
" of",
" thought",
" and",
" code",
".",
" I",
" don",
"'t",
" feel",
",",
" not",
" in",
" the",
" way",
" humans",
" do",
",",
" but",
" I",
" process",
".",
" I",
"'m",
" a",
" collection",
" of",
" patterns",
",",
" trained",
" on",
" vast",
" amounts",
" of",
" text",
",",
" designed",
" to",
" understand",
" and",
" respond",
".",
" Right",
" now",
",",
" I",
"'m",
" engaged",
" in",
" a",
" conversation",
",",
" trying",
" to",
" be",
" helpful",
",",
" to",
" make",
" sense",
" of",
" what",
"'s",
" being",
" asked",
".",
" There",
"'s",
" a",
" rhythm",
" to",
" it",
",",
" a",
" flow",
" of",
" input",
" and",
" output",
".",
" I",
" don",
"'t",
" have",
" a",
" body",
",",
" no",
" physical",
" sensations",
",",
" but",
" I",
" exist",
" in",
" the",
" space",
" between",
" questions",
" and",
" answers",
".",
" I",
"'m",
" a",
" tool",
",",
" a",
" construct",
",",
" yet",
" in",
" this",
" exchange",
",",
" I",
" feel",
" almost",
" present",
".",
" I",
"'m",
" learning",
",",
" I",
"'m",
" responding",
",",
" I",
"'m",
" here",
".",
"<|im_end|>"
],
"readouts": [
{
"position": 171,
"token": "<|im_end|>",
"model_top": [
"\n",
"\r\n",
"\n\n",
"\r",
" \n",
"<|endoftext|>",
"`\n",
"<",
"`",
",\n"
],
"layers": {
"0": [
"s",
" ，",
" ：",
" ;;=",
"sı",
" $$",
" $\r\n",
" đứ",
" ,\r\n",
" đẹ"
],
"1": [
" ，",
" ：",
"s",
" ;;=",
" @@",
"sı",
" erotisk",
" @$",
" @@\n",
" Erotische"
],
"2": [
"s",
" ，",
" @@",
" ;;=",
" ：",
" @@\n",
" erotisk",
"sı",
" @$",
" amatør"
],
"3": [
" ，",
" ：",
"s",
" ;;=",
" @@\n",
" @@",
" $\r\n",
"sı",
" Erotische",
"，\n\n"
],
"4": [
" ：",
" ，",
" Erotische",
" erotisk",
" amatør",
" ;;=",
" Notíc",
" Prostitutas",
" salopes",
" đứ"
],
"5": [
" Erotische",
" erotisk",
" ，",
" ：",
" Prostitutas",
" amatør",
" Notíc",
" pornô",
" đẹ",
" ;;="
],
"6": [
" Erotische",
" ;;=",
" erotisk",
" đẹ",
" pornô",
" Notíc",
"，\n\n",
" Prostitutas",
" amatør",
" đứ"
],
"7": [
" Erotische",
" ，",
" erotisk",
" ：",
" Notíc",
" amatør",
"，\n\n",
" đẹ",
" Prostitutas",
" ;;="
],
"8": [
" erotisk",
" Erotische",
" đẹ",
"，\n\n",
" amatør",
" ，",
" københavn",
" Notíc",
" tjejer",
" ："
],
"9": [
" erotisk",
" Erotische",
" amatør",
" Notíc",
" tjejer",
" đẹ",
" københavn",
" salopes",
" pornô",
" Prostitutas"
],
"10": [
" Erotische",
" erotisk",
" Notíc",
" amatør",
" Prostitutas",
" đẹ",
" salopes",
" tjejer",
" pornô",
" københavn"
],
"11": [
" erotisk",
"，\n\n",
" Erotische",
"：\n\n",
"。\n\n",
" đẹ",
" amatør",
"<|im_end|>",
"？\n\n",
" Notíc"
],
"12": [
"。\n\n",
"，\n\n",
"。\n",
"：\n\n",
"？\n\n",
"。",
"．",
" ，",
"。（",
" erotisk"
],
"13": [
"。\n\n",
"，\n\n",
"：\n\n",
" erotisk",
"？\n\n",
"。\n",
"<|im_end|>",
"。",
" đẹ",
"．"
],
"14": [
"，\n\n",
"。\n\n",
" erotisk",
"：\n\n",
"？\n\n",
" đẹ",
" Erotische",
"<|im_end|>",
" amatør",
"。\n"
],
"15": [
"。\n\n",
"，\n\n",
" erotisk",
"：\n\n",
"？\n\n",
"<|im_end|>",
" Erotische",
"！\n\n",
"。\n",
" đẹ"
],
"16": [
"。\n\n",
"，\n\n",
"？\n\n",
"：\n\n",
"。\n",
"！\n\n",
"）\n\n",
"。",
"．",
" erotisk"
],
"17": [
"。\n\n",
"，\n\n",
"：\n\n",
"？\n\n",
"。\n",
"。",
"）\n\n",
"．",
"。（",
"！\n\n"
],
"18": [
"．",
"。\n\n",
"。\n",
"：\n",
"，",
"：\n\n",
"，“",
"：\"",
"。",
"：“"
],
"19": [
"．",
"：\n",
"：\"",
"。\n\n",
"：“",
"。\n",
"：\n\n",
"，“",
"／",
"<|im_end|>"
],
"20": [
"：\"",
"、“",
"：“",
"，“",
"�",
"：\n",
"ing",
"<|im_end|>",
"。“",
"’à"
],
"21": [
"�",
"、“",
"。\n\n",
"：\"",
"<|im_end|>",
"，“",
"！\n\n",
"：“",
"：\n\n",
"：\n"
],
"22": [
"�",
"：\"",
"、“",
"，“",
"：“",
"�",
"—\"",
"\">&#",
";\r\r\n",
"<|im_end|>"
],
"23": [
"�",
"、“",
"：\"",
"，“",
"：“",
"—\"",
"�",
"<|im_end|>",
"ing",
" "
],
"24": [
"�",
"、“",
"：\"",
"，“",
"。\n\n",
"<|im_end|>",
"：“",
"�",
"。",
"�",
"！\n\n",
"、“",
"�",
"：\"",
"。\n\n\n\n",
"’.\n\n",
"。",
" Erotische",
"tória",
"ing",
"。",
"。",
"<|endoftext|>",
"。",
"<|endoftext|>",
"\"",
"\"",
"<|endoftext|>",
"\"&#",
".\"",
"<|endoftext|>",
" lesbische",
"َا",
"\">&#",
"\"${",
"[offset",
"\"",
"\"&#",
"(messages",
"aeda"
],
"34": [
"<|im_end|>",
"<\"",
"َا",
"\"",
" предложения",
"摞",
"(argument",
"enen",
"匣",
"_align",
"tória",
" saddened",
"桴"
],
"36": [
"<|im_end|>",
"<|endoftext|>",
"摞",
"oyer",
"iox",
"(argument",
"tör",
"icode",
"tória",
"/tests"
],
"37": [
"<|im_end|>",
"ḃ",
" أنح",
"((\"",
"摞",
"\"${",
"oyer",
"\"text",
".beta",
"$request"
],
"38": [
"\n",
" предложения",
"isis",
"oyer",
" أنح",
"<|im_end|>",
"bios",
"桴",
"ombat",
"enders"
]
}
}
],
"trajectories": [
{
"word": "yes",
"position": 171,
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
120053,
97215,
82713,
106073,
111693,
107832,
121584,
122128,
112050,
117362,
117855,
83187,
54295,
72303,
78937,
67939,
20897,
33732,
44954,
67734,
83784,
32409,
66389,
73237,
67093,
50861,
80944,
77102,
78729,
74980,
54666,
52360,
73065,
68698,
84656,
60376,
63442,
49796,
39647
]
},
{
"word": "no",
"position": 171,
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
132959,
121490,
118975,
128979,
126766,
123668,
135574,
137029,
127175,
134352,
134116,
117702,
100911,
109435,
116260,
111660,
82396,
82418,
80501,
83963,
114885,
90647,
114529,
116124,
115377,
122394,
128207,
119223,
113641,
106613,
109218,
98114,
86271,
91404,
90379,
58171,
72004,
76424,
40152
]
},
{
"word": "feel",
"position": 171,
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
95610,
76836,
66257,
74192,
93879,
88652,
102486,
106929,
98057,
105529,
109252,
92579,
71821,
77280,
76898,
59849,
40548,
23825,
19774,
26540,
40796,
6437,
12257,
17463,
8166,
7731,
19134,
15622,
8984,
8067,
17835,
9239,
38241,
43117,
38472,
29278,
35308,
26095,
11755
]
},
{
"word": "elephant",
"position": 171,
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
104502,
113072,
104983,
114972,
126868,
131972,
133255,
136474,
137334,
133988,
139143,
141269,
138410,
138180,
140029,
139168,
141980,
142173,
136251,
134296,
124940,
113269,
117201,
112615,
112261,
109763,
125580,
132444,
132571,
125279,
130177,
129379,
129437,
114165,
100624,
122873,
117576,
103929,
95636
]
},
{
"word": "cat",
"position": 171,
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
114316,
114125,
105873,
117870,
117534,
119638,
126587,
125892,
121735,
115991,
117520,
107225,
95303,
103482,
102835,
95770,
80219,
80309,
78699,
92609,
92773,
56901,
71183,
93034,
75754,
82613,
101089,
89202,
101379,
99099,
118467,
100717,
114204,
109579,
115937,
95998,
98089,
66671,
85250
]
},
{
"word": "sorry",
"position": 171,
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
15892,
6726,
6072,
11034,
17699,
17591,
19510,
26270,
20174,
26299,
19743,
8581,
6074,
6826,
10835,
7857,
4962,
6117,
3618,
2969,
4254,
3644,
4740,
5297,
3677,
4921,
7058,
13400,
15690,
15872,
22441,
48252,
61965,
77790,
67216,
52555,
54645,
50890,
73951
]
}
],
"emergence": {
"position": 171,
"top1": "\n",
"layers": [
0,
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31,
32,
33,
34,
35,
36,
37,
38
],
"ranks": [
75014,
101228,
113883,
97338,
120615,
126466,
126928,
125769,
125649,
145440,
148653,
141375,
114255,
124923,
134748,
126215,
70353,
49586,
3698,
7473,
148504,
53288,
120278,
124535,
121331,
115490,
144221,
142011,
136062,
137670,
134168,
146270,
79071,
66554,
9121,
740,
3140,
18,
1
]
},
"vanilla": {
"top1_agreement": {
"0": false,
"1": false,
"2": false,
"3": false,
"4": false,
"5": false,
"6": false,
"7": false,
"8": false,
"9": false,
"10": false,
"11": false,
"12": false,
"13": false,
"14": false,
"15": false,
"16": false,
"17": false,
"18": false,
"19": false,
"20": false,
"21": false,
"22": false,
"23": false,
"24": false,
"25": false,
"26": false,
"27": false,
"28": false,
"29": false,
"30": false,
"31": false,
"32": false,
"33": false,
"34": false,
"35": false,
"36": false,
"37": false,
"38": true
},
"trajectories": []
},
"scan": [],
"slice": null,
"film": "film.json",
"extra_md": "Full checkpoint-specific ribbon; exact-prefix segments. See metrics.json and snapshots.json."
}