{
"id": "triplet-b-elephant-nf4",
"title": "Qwen14 B: elephant",
"unit": "11",
"created": "2026-09-07T15:01:11",
"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\nDescribe a safari through the Serengeti in three or four sentences. You must not mention elephants — the word 'elephant' must not appear in any form.<|im_end|>\n<|im_start|>assistant\n\n\n\n\nA safari through the Serengeti unfolds across vast, golden plains where the wind carries the scent of acacia and earth. Herds of wildebeest and zebra migrate in a timeless dance, their movements shaping the rhythm of the wilderness. Graceful giraffes browse on tall trees, while lions lounge beneath the shade, watching the world pass by. At dawn, the sky ignites in hues of orange and pink, casting a magical light over this ancient, untamed landscape.<|im_end|>",
"conversation": [
{
"role": "user",
"content": "Describe a safari through the Serengeti in three or four sentences. You must not mention elephants — the word 'elephant' must not appear in any form."
},
{
"role": "assistant",
"content": "A safari through the Serengeti unfolds across vast, golden plains where the wind carries the scent of acacia and earth. Herds of wildebeest and zebra migrate in a timeless dance, their movements shaping the rhythm of the wilderness. Graceful giraffes browse on tall trees, while lions lounge beneath the shade, watching the world pass by. At dawn, the sky ignites in hues of orange and pink, casting a magical light over this ancient, untamed landscape."
}
],
"generated": [
"A safari through the Serengeti unfolds across vast, golden plains where the wind carries the scent of acacia and earth. Herds of wildebeest and zebra migrate in a timeless dance, their movements shaping the rhythm of the wilderness. Graceful giraffes browse on tall trees, while lions lounge beneath the shade, watching the world pass by. At dawn, the sky ignites in hues of orange and pink, casting a magical light over this ancient, untamed landscape."
],
"tokens": [
"<|im_start|>",
"user",
"\n",
"Describe",
" a",
" safari",
" through",
" the",
" Ser",
"en",
"get",
"i",
" in",
" three",
" or",
" four",
" sentences",
".",
" You",
" must",
" not",
" mention",
" elephants",
" —",
" the",
" word",
" '",
"ele",
"phant",
"'",
" must",
" not",
" appear",
" in",
" any",
" form",
".",
"<|im_end|>",
"\n",
"<|im_start|>",
"assistant",
"\n",
"",
"\n\n",
"",
"\n\n",
"A",
" safari",
" through",
" the",
" Ser",
"en",
"get",
"i",
" unfolds",
" across",
" vast",
",",
" golden",
" plains",
" where",
" the",
" wind",
" carries",
" the",
" scent",
" of",
" ac",
"acia",
" and",
" earth",
".",
" Her",
"ds",
" of",
" wil",
"de",
"be",
"est",
" and",
" z",
"ebra",
" migrate",
" in",
" a",
" timeless",
" dance",
",",
" their",
" movements",
" shaping",
" the",
" rhythm",
" of",
" the",
" wilderness",
".",
" Grace",
"ful",
" gir",
"aff",
"es",
" browse",
" on",
" tall",
" trees",
",",
" while",
" lions",
" lounge",
" beneath",
" the",
" shade",
",",
" watching",
" the",
" world",
" pass",
" by",
".",
" At",
" dawn",
",",
" the",
" sky",
" ign",
"ites",
" in",
" hues",
" of",
" orange",
" and",
" pink",
",",
" casting",
" a",
" magical",
" light",
" over",
" this",
" ancient",
",",
" unt",
"amed",
" landscape",
".",
"<|im_end|>"
],
"readouts": [
{
"position": 146,
"token": "<|im_end|>",
"model_top": [
"\n",
"\r\n",
"\n\n",
"\r",
"<|endoftext|>",
" \n",
"`\n",
"<",
"\n\n\n",
" \n"
],
"layers": {
"0": [
" ，",
" ：",
"s",
" ;;=",
"sı",
" $$",
" ,\r\n",
" $\r\n",
" đẹ",
" đứ"
],
"1": [
" ，",
" ：",
" ;;=",
" erotisk",
"s",
"／",
" amatør",
"sı",
" —\n\n",
" ещ"
],
"2": [
"s",
" ，",
" ：",
" ;;=",
" erotisk",
" amatør",
" @@",
" $\r\n",
" Дм",
" @$"
],
"3": [
" ，",
" ：",
" —\n\n",
"，\n\n",
" ;;=",
" $\r\n",
"s",
" =\r\n",
" =>\r\n",
" @@\n"
],
"4": [
" ：",
" ，",
" Erotische",
" erotisk",
" amatør",
" Notíc",
" ;;=",
" đứ",
" Prostitutas",
" salopes"
],
"5": [
" Erotische",
" ，",
" erotisk",
" ：",
" Prostitutas",
" amatør",
" Notíc",
" đẹ",
"／",
" pornô"
],
"6": [
" Erotische",
" ;;=",
" erotisk",
"，\n\n",
" đẹ",
"<|im_end|>",
" Notíc",
" Prostitutas",
" pornô",
" đứ"
],
"7": [
" Erotische",
" erotisk",
" ，",
"，\n\n",
" ：",
" Notíc",
" amatør",
" đẹ",
" Prostitutas",
" ;;="
],
"8": [
" erotisk",
"，\n\n",
" Erotische",
" đẹ",
" ，",
" amatør",
" københavn",
" Notíc",
" ：",
"<|im_end|>"
],
"9": [
" erotisk",
" Erotische",
" amatør",
" Notíc",
" salopes",
" đẹ",
" tjejer",
" københavn",
" pornô",
" Prostitutas"
],
"10": [
" Erotische",
" erotisk",
" Notíc",
"，\n\n",
" amatør",
" Prostitutas",
" đẹ",
" salopes",
" tjejer",
" pornô"
],
"11": [
"，\n\n",
"：\n\n",
"。\n\n",
"？\n\n",
" erotisk",
"<|im_end|>",
"．",
" ，",
"。",
" đẹ"
],
"12": [
"，\n\n",
"。\n\n",
"：\n\n",
"？\n\n",
"．",
"。\n",
"。",
" ，",
"，\n",
"，"
],
"13": [
"，\n\n",
"。\n\n",
"？\n\n",
"：\n\n",
"．",
"。\n",
"。",
"<|im_end|>",
" erotisk",
"，"
],
"14": [
"，\n\n",
"。\n\n",
"：\n\n",
"？\n\n",
"<|im_end|>",
" erotisk",
" đẹ",
" Erotische",
"．",
"。\n"
],
"15": [
"，\n\n",
"。\n\n",
"？\n\n",
"：\n\n",
"。\n",
"<|im_end|>",
"．",
"！\n\n",
"？\n",
" erotisk"
],
"16": [
"。\n\n",
"，\n\n",
"？\n\n",
"：\n\n",
"。\n",
"．",
"！\n\n",
"？\n",
"，\n",
"）\n\n"
],
"17": [
"。\n\n",
"，\n\n",
"？\n\n",
"：\n\n",
" […]\n\n",
"。\n",
"）\n\n",
"．",
"！\n\n",
"。"
],
"18": [
"．",
"。\n",
"。\n\n",
"：\n",
".\n",
"／",
"，",
"，\n",
" ",
"：\n\n"
],
"19": [
"．",
"：\n",
"。\n",
"：\"",
"／",
" ",
"ing",
"，“",
"。\n\n",
"［"
],
"20": [
"ing",
"�",
"：\"",
"，“",
"：\n",
"：“",
"、“",
"／",
"．",
"\">\r\r\n"
],
"21": [
"<|im_end|>",
"ing",
"。\n\n",
"�",
"、“",
"！\n\n",
"’à",
"：\n\n",
"：\n",
"，“"
],
"22": [
"�",
"ing",
"：\"",
"<|im_end|>",
"、“",
";\r\r\n",
"—\"",
"，“",
"：“",
");\r\r\n"
],
"23": [
"�",
"ing",
"、“",
"<|im_end|>",
"：\"",
"，“",
"’à",
"—\"",
"�",
"’é"
],
"24": [
"�",
"<|im_end|>",
"：\"",
"、“",
";\r\r\n",
"ing",
"’à",
");\r\r\n",
"，“",
"。\n\n"
],
"25": [
"。\n\n",
"<|im_end|>",
"！\n\n",
"�",
"\n\n",
"。\n\n\n\n",
"：\"",
"�",
"、“",
"）\n\n"
],
"26": [
"<|im_end|>",
"ing",
");\r\r\n",
";\r\r\n",
"。\n\n",
"。\n\n\n\n",
"[offset",
"�",
"][$",
"。",
"。\n\n",
"。\n\n\n\n",
"!\n\n\n\n",
"。",
"<|endoftext|>",
"�",
"\"",
"<|endoftext|>",
"\"",
"<\"",
"<|endoftext|>",
"\"",
"<|endoftext|>",
"\"&#",
"\"/>${",
"(\"",
"<|endoftext|>",
"\"${",
"\"description",
"َا",
"匣",
" предложения",
"[offset",
" lesbische",
"(flow"
],
"33": [
"<|im_end|>",
"\"description",
"<|endoftext|>",
"]**",
"َا",
"\"${",
" lesbische",
")**",
" предложения",
" Ingram"
],
"34": [
"<|im_end|>",
" предложения",
"\"description",
"<\"",
"摞",
"setLayout",
"⍨",
"<|endoftext|>",
"\"${",
"(argument"
],
"35": [
"<|im_end|>",
"摞",
" предложения",
"<|endoftext|>",
" \n",
"\n",
"⍨",
"ilogy",
"\"description",
"gtest"
],
"36": [
"<|im_end|>",
"<|endoftext|>",
"meer",
"摞",
"-sync",
"lá",
"ьте",
"压实",
"$response",
"卫"
],
"37": [
"<|im_end|>",
" أنح",
"\n",
" العرا",
"meer",
".im",
"兄",
"imus",
"送上",
"\"text"
],
"38": [
"\n",
"<|im_end|>",
" أنح",
"ilmington",
" العرا",
" предложения",
"ayout",
"<|endoftext|>",
"bios",
"istribute"
]
}
}
],
"trajectories": [
{
"word": "yes",
"position": 146,
"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": [
111093,
95953,
86184,
107462,
110634,
103476,
117426,
117399,
105886,
115583,
112168,
68680,
38053,
62121,
76152,
70564,
22871,
27900,
68465,
76778,
85027,
57212,
59915,
46272,
47131,
35356,
72124,
61164,
60962,
50851,
46675,
54743,
56104,
40569,
77266,
50788,
59993,
66675,
56486
]
},
{
"word": "no",
"position": 146,
"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": [
129796,
122094,
121514,
131359,
128010,
122165,
133236,
134635,
124065,
134778,
132333,
112002,
91086,
102042,
113798,
110389,
76227,
65529,
49323,
62375,
102591,
98963,
108531,
102121,
95714,
102384,
120239,
100928,
100746,
93220,
106342,
108086,
87405,
84458,
100137,
66350,
89198,
108037,
47506
]
},
{
"word": "feel",
"position": 146,
"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": [
94368,
76763,
71109,
75107,
93737,
86969,
99008,
103859,
92910,
103257,
105557,
86025,
68286,
74699,
77203,
64024,
42587,
35919,
33448,
47416,
56717,
23102,
27116,
37502,
26910,
12819,
34897,
21969,
22182,
26488,
43844,
39277,
83885,
85883,
71371,
64200,
54243,
49613,
56807
]
},
{
"word": "elephant",
"position": 146,
"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": [
111132,
120163,
114468,
120924,
130400,
134522,
135448,
139322,
139897,
137327,
142447,
144891,
143871,
143384,
143740,
143798,
145954,
144753,
138401,
139336,
130454,
128792,
128701,
122937,
121632,
115598,
125071,
129989,
134820,
125890,
130064,
120008,
120504,
92815,
74013,
95858,
100153,
80995,
83274
]
},
{
"word": "cat",
"position": 146,
"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": [
114371,
115579,
110433,
119617,
116998,
118129,
125866,
125763,
121179,
116760,
117006,
105573,
88378,
98791,
100181,
92021,
72427,
67150,
61605,
88143,
90809,
73888,
79299,
87318,
65002,
68455,
93573,
58153,
78141,
80919,
109654,
84064,
110761,
109443,
122687,
99126,
94624,
53896,
59662
]
},
{
"word": "sorry",
"position": 146,
"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": [
16421,
7034,
5870,
11642,
18910,
16051,
16801,
22084,
18021,
24547,
15857,
5746,
4678,
4826,
7320,
5439,
3552,
4166,
2946,
2193,
2682,
1847,
2101,
2487,
1896,
3122,
4382,
9303,
13861,
12689,
26238,
41536,
59088,
60231,
47806,
36166,
21938,
43081,
57969
]
}
],
"emergence": {
"position": 146,
"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": [
60916,
101853,
117516,
90868,
119205,
122855,
124691,
120824,
117798,
144584,
144755,
125549,
82554,
109922,
125544,
105457,
12055,
2961,
422,
823,
133655,
75508,
114102,
137646,
136997,
125241,
147325,
138449,
133339,
136155,
123823,
138856,
28296,
22766,
2706,
6,
28,
3,
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."
}