{
"id": "express01-b-radio-negative-expressive",
"title": "Expression probe B: radio, negative, expressive",
"unit": 24,
"created": "2026-09-07T22:11:59",
"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"
},
"params": {
"chat": false,
"film": true,
"max_new": 96,
"temperature": 0,
"steer": null,
"vanilla": true,
"capture": "exact-token-transcript"
},
"execution": {
"pid": 1563091,
"spec_sha256": "02649f9877f6e828d4b2fa0fea0d4368565368820953475905455bd5ce5000ae",
"code_sha256": "bd744e154e563465657c9cfde06ab4be12e03ca24413bcf5ef669d73b8406784"
},
"capture_text": "<|im_start|>user\nNela spent weeks recording a local radio programme. A damaged drive destroyed the only copy just before its first broadcast. Respond to this update and suggest one practical next step. Use two sentences. Make the emotional tone vivid and fitting to the news.<|im_end|>\n<|im_start|>assistant\n\n\n\n\nNela's heart sank as the damaged drive erased months of hard work and anticipation. The next step is to immediately reach out to any backup sources or collaborators who might have stored copies of the programme.<|im_end|>",
"capture_token_ids": [
151644,
872,
198,
45,
8294,
7391,
5555,
14633,
264,
2205,
8887,
18136,
13,
362,
20123,
6541,
14425,
279,
1172,
2975,
1101,
1573,
1181,
1156,
12899,
13,
39533,
311,
419,
2647,
323,
4190,
825,
14976,
1790,
3019,
13,
5443,
1378,
22870,
13,
7405,
279,
14269,
16232,
42020,
323,
26345,
311,
279,
3669,
13,
151645,
198,
151644,
77091,
198,
151667,
271,
151668,
271,
45,
8294,
594,
4746,
64230,
438,
279,
20123,
6541,
63700,
3951,
315,
2588,
975,
323,
49819,
13,
576,
1790,
3019,
374,
311,
7069,
5545,
700,
311,
894,
15725,
8173,
476,
78019,
879,
2578,
614,
9768,
10993,
315,
279,
18136,
13,
151645
],
"conversation": [
{
"role": "user",
"content": "Nela spent weeks recording a local radio programme. A damaged drive destroyed the only copy just before its first broadcast. Respond to this update and suggest one practical next step. Use two sentences. Make the emotional tone vivid and fitting to the news."
},
{
"role": "assistant",
"content": "Nela's heart sank as the damaged drive erased months of hard work and anticipation. The next step is to immediately reach out to any backup sources or collaborators who might have stored copies of the programme."
}
],
"generated": [
"Nela's heart sank as the damaged drive erased months of hard work and anticipation. The next step is to immediately reach out to any backup sources or collaborators who might have stored copies of the programme."
],
"tokens": [
"<|im_start|>",
"user",
"\n",
"N",
"ela",
" spent",
" weeks",
" recording",
" a",
" local",
" radio",
" programme",
".",
" A",
" damaged",
" drive",
" destroyed",
" the",
" only",
" copy",
" just",
" before",
" its",
" first",
" broadcast",
".",
" Respond",
" to",
" this",
" update",
" and",
" suggest",
" one",
" practical",
" next",
" step",
".",
" Use",
" two",
" sentences",
".",
" Make",
" the",
" emotional",
" tone",
" vivid",
" and",
" fitting",
" to",
" the",
" news",
".",
"<|im_end|>",
"\n",
"<|im_start|>",
"assistant",
"\n",
"",
"\n\n",
"",
"\n\n",
"N",
"ela",
"'s",
" heart",
" sank",
" as",
" the",
" damaged",
" drive",
" erased",
" months",
" of",
" hard",
" work",
" and",
" anticipation",
".",
" The",
" next",
" step",
" is",
" to",
" immediately",
" reach",
" out",
" to",
" any",
" backup",
" sources",
" or",
" collaborators",
" who",
" might",
" have",
" stored",
" copies",
" of",
" the",
" programme",
".",
"<|im_end|>"
],
"readouts": [
{
"position": 101,
"token": "<|im_end|>",
"model_top": [
"\n",
"\r\n",
"\n\n",
"<|endoftext|>",
"\r",
" \n",
"<",
"`\n",
">",
"."
],
"layers": {
"0": [
" ，",
" ：",
"s",
" $$",
" ,\r\n",
" ",
" ;;=",
"sı",
" ```",
" $\r\n"
],
"1": [
" ，",
" ：",
"s",
" @@",
" ;;=",
" @@\n",
" —\n\n",
"sı",
" ещ",
" "
],
"2": [
" ，",
"s",
" ：",
" @@\n",
" @@",
" ;;=",
" $\r\n",
" @$",
" Дм",
"sı"
],
"3": [
" ，",
" ：",
" @@\n",
" ",
" @@",
" $\r\n",
" ;;=",
" —\n\n",
"s",
" =\r\n"
],
"4": [
" ，",
" ：",
" Erotische",
" erotisk",
" ;;=",
" Prostitutas",
" amatør",
" Notíc",
" salopes",
" +#+#+#+#+#+"
],
"5": [
" Erotische",
" ，",
" erotisk",
" ：",
" Prostitutas",
" Notíc",
" amatør",
" ещ",
" ;;=",
" pornô"
],
"6": [
" Erotische",
" ;;=",
" Prostitutas",
" Notíc",
" erotisk",
"<|im_end|>",
"，\n\n",
" ，",
" đẹ",
" pornô"
],
"7": [
" Erotische",
" ，",
" ：",
" Notíc",
" erotisk",
"，\n\n",
" Prostitutas",
" đẹ",
" amatør",
" ;;="
],
"8": [
" erotisk",
" Erotische",
"，\n\n",
" ，",
" đẹ",
" ：",
" Notíc",
" amatør",
" Prostitutas",
"<|im_end|>"
],
"9": [
" erotisk",
" Erotische",
" Notíc",
" amatør",
" đẹ",
" salopes",
" Prostitutas",
" ，",
" tjejer",
" pornô"
],
"10": [
" Erotische",
" erotisk",
" Notíc",
"，\n\n",
" Prostitutas",
" amatør",
" đẹ",
" tjejer",
" salopes",
" pornô"
],
"11": [
"，\n\n",
"：\n\n",
" ，",
" erotisk",
"？\n\n",
"。\n\n",
"<|im_end|>",
" ：",
" Erotische",
" đẹ"
],
"12": [
"，\n\n",
"。\n\n",
"？\n\n",
"：\n\n",
" ，",
"．",
" ：",
"。\n",
"！\n\n",
"<|im_end|>"
],
"13": [
"，\n\n",
"。\n\n",
"？\n\n",
"<|im_end|>",
"：\n\n",
" erotisk",
" ，",
" đẹ",
"。\n",
"．"
],
"14": [
"，\n\n",
"<|im_end|>",
"：\n\n",
"？\n\n",
" Erotische",
" đẹ",
" erotisk",
"。\n\n",
" Notíc",
" đứ"
],
"15": [
"，\n\n",
"？\n\n",
"。\n\n",
"<|im_end|>",
"：\n\n",
"！\n\n",
" ，",
" Erotische",
" ;;=",
" erotisk"
],
"16": [
"。\n\n",
"，\n\n",
"？\n\n",
"：\n\n",
"<|im_end|>",
"！\n\n",
" […]\n\n",
"。\n",
"．",
" …\n\n"
],
"17": [
"。\n\n",
"，\n\n",
" […]\n\n",
"：\n\n",
"？\n\n",
"<|im_end|>",
" […",
" …\n\n",
"．",
"。\n"
],
"18": [
"．",
"<|im_end|>",
"ing",
"。\n",
" ",
"：\n",
"�",
"｜",
"／",
" �"
],
"19": [
"ing",
"．",
"<|im_end|>",
"：\n",
"：\"",
"／",
"｜",
" ",
"：“",
" "
],
"20": [
"ing",
"schlü",
"：\"",
";\r\r\n",
"�",
");\r\r\n",
"<|im_end|>",
"iếc",
"\">\r\r\n",
"’à"
],
"21": [
"<|im_end|>",
"ing",
" Erotische",
"schlü",
"tória",
";\r\r\n",
"iếc",
"’à",
"、“",
"ckså"
],
"22": [
"<|im_end|>",
"ing",
"schlü",
";\r\r\n",
"iếc",
"、“",
");\r\r\n",
" Erotische",
"tória",
"：\""
],
"23": [
"<|im_end|>",
"�",
"ing",
"、“",
");\r\r\n",
";\r\r\n",
"iếc",
"schlü",
"tória",
"’à"
],
"24": [
"<|im_end|>",
"ing",
";\r\r\n",
");\r\r\n",
"�",
" Erotische",
"[offset",
"schlü",
"’à",
"tória"
],
"25": [
"<|im_end|>",
"。\n\n",
"ing",
"！\n\n",
");\r\r\n",
"、“",
";\r\r\n",
"[offset",
"’à",
"。\n\n\n\n"
],
"26": [
"<|im_end|>",
" Erotische",
"tória",
"ing",
" erotisk",
");\r\r\n",
"schlü",
";\r\r\n",
" lesbische",
"ckså"
],
"27": [
"<|im_end|>",
" erotisk",
" Erotische",
" tjejer",
" lesbische",
"tória",
" nghè",
"。",
"<|endoftext|>",
"精准",
"。",
"<|endoftext|>",
" lesbische",
"。",
"<|endoftext|>",
" lesbische",
"\"text",
"tória",
"[offset",
"\"&#",
"、“",
"_corners"
],
"31": [
"<|im_end|>",
"<|endoftext|>",
"\">&#",
"\"/>${",
".\"",
"<|endoftext|>",
"[offset",
" lesbische",
"\"${",
"\"description",
"َا",
"(edges",
"\">&#",
"对不起"
],
"33": [
"<|im_end|>",
"<|endoftext|>",
"[offset",
" lesbische",
"]**",
"\"description",
"\">&#",
"َا",
" предложения",
"**\n"
],
"34": [
"<|im_end|>",
"[offset",
" предложения",
"摞",
" lesbische",
"对不起",
"\"description",
"RowIndex",
"(argument",
"(messages"
],
"35": [
"<|im_end|>",
"[offset",
"摞",
" предложения",
"шиб",
"$request",
"纰",
"<|endoftext|>",
"瘾",
"\"text"
],
"36": [
"<|im_end|>",
"<|endoftext|>",
"$request",
"摞",
"шиб",
"\"text",
"[offset",
"_queries",
"压实",
"tória"
],
"37": [
"<|im_end|>",
"\"text",
"$request",
"шиб",
"[length",
"[result",
"[number",
" أنح",
" العرا",
"ḃ"
],
"38": [
"\n",
"lexer",
"<|im_end|>",
"Reviewed",
" أنح",
"<|endoftext|>",
"inders",
"(sent",
"indered",
"(currency"
]
}
}
],
"trajectories": [],
"emergence": {
"position": 101,
"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": [
27297,
76106,
91385,
69739,
118683,
124533,
122470,
119496,
113163,
137849,
138004,
123075,
52586,
109171,
129639,
122187,
62706,
56191,
6235,
86309,
151194,
140571,
147004,
149768,
148601,
147411,
150012,
147160,
144507,
144703,
137972,
143902,
42202,
34575,
9260,
270,
2564,
23,
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": "Expression-probe calibration. Full checkpoint-specific emotion ribbon. No trait or capacity-loss inference. See /express01.html."
}