layer int64 | feature_index int64 | description string | explanation string | top_examples list | top_examples_tks list | average_activation float64 | top_activating_tokens list | raw_explanation string | language_distribution dict | general_language_distribution dict | filename string |
|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 0 | Unknown | No explanation generated | [
"<<personal>> ist immer bemüht alles wieder in Ordnung zu bringen. Die sind auch immer",
" je von Faktoren wie kompetentem und verlässlichem Führungs <<personal>> und Qualität in",
"sstelle mitzuteilen. Ihre Briefpost wird Ihnen täglich vom Pflege <<personal>> ausgehänd",
" der über den Web-Dienst bereitge... | [
{
"activations": [
0,
12.1328125,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 12.1328125,
"tokens": [
15782,
37030,
742,
3135,
58364,
6089,
2900,
... | 7.209713 | [
{
"average_activation": 7.98578125,
"frequency": 50,
"token": " Personal",
"token_id": 10743
},
{
"average_activation": 11.751953125,
"frequency": 8,
"token": "personal",
"token_id": 37030
},
{
"average_activation": 7.403459821428571,
"frequency": 7,
"token": " ال... | {
"arb_Arab": 0.41,
"cmn_Hani": 0.02,
"deu_Latn": 0.57,
"fra_Latn": null
} | {
"arb_Arab": 0.5837837837837838,
"cmn_Hani": 0.010810810810810811,
"deu_Latn": 0.3567567567567568,
"fra_Latn": 0.04864864864864865
} | feature_0_complete.json | |
0 | 10,000 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10000_complete.json | |
0 | 10,001 | Unknown | No explanation generated | [
" الديكتاتورية الإيرانية القاتلة في طهران العقبة الرئيسية أمام السلام والاستقرار",
"<<فير الامريكي>> في\" نواكشوط\" :\" امريكا ترغب في اجراء حوار مثمر",
"<<الامريكي>> في اكتوبر من عام 2016، على خلفية أحداث الحادي عشر من سبتمبر حول",
"قود الاجلة لبرنت والخ <<ام الامريكي>> تهبط أكثر من دولارين نيويورك",
"<... | [
{
"activations": [
7.9765625,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 7.9765625,
"tokens": [
54531,
117995,
2632,
413,
4494,
27310,
1018,
... | 7.528716 | [
{
"average_activation": 7.454963235294118,
"frequency": 17,
"token": " اي",
"token_id": 8892
},
{
"average_activation": 7.6515625,
"frequency": 10,
"token": " الامريكي",
"token_id": 54531
},
{
"average_activation": 7.46875,
"frequency": 3,
"token": " ١٤",
"tok... | {
"arb_Arab": 1,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": null
} | {
"arb_Arab": 1,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": null
} | feature_10001_complete.json | |
0 | 10,002 | Unknown | No explanation generated | [
"Architektur ihre Stärken aus, wenn hohe Single <<-Thread-Performance.>> Und",
" einen Glanz <<turnhosenschwanzabzeichner-Thread.>> Ich kann",
"une […] Avec Penny D <<readful,>> John Logan a eu l’idée",
".ocx 文件）。运行 <<read>> .exe软件，点击任意键",
" zu glauben, aber selbst RES kann aus memeverseuchten Pun <<-Thr... | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
11.6640625,
0,
0,
0,
0
],
"max_val": 11.6640625,
"tokens": [
17,
103416,
3446,
68897,
677,
16,
3216,
157... | 11.532813 | [
{
"average_activation": 11.5328125,
"frequency": 5,
"token": "read",
"token_id": 3730
}
] | {
"arb_Arab": null,
"cmn_Hani": 0.2,
"deu_Latn": 0.6,
"fra_Latn": 0.2
} | {
"arb_Arab": null,
"cmn_Hani": 0.2,
"deu_Latn": 0.6,
"fra_Latn": 0.2
} | feature_10002_complete.json | |
0 | 10,003 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10003_complete.json | |
0 | 10,004 | Unknown | No explanation generated | [
" roten Teppich. Bei den Crit <<ics‘ Choice>> Awards zeigte sie, wie gut",
" 8. Isolated Through Multiplicity 9. Individual <<Choice>> 10. Glimps",
"<<Choice>> Hotels, Hotels.com و Booking.com بحيث يمكنك القيام بحجوزاتك في"
] | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
11.984375,
0,
0,
0,
0,
0,
0
],
"max_val": 11.984375,
"tokens": [
730,
43266,
54452,
18,
6281,
634,
35096,
1732... | 11.684896 | [
{
"average_activation": 11.684895833333334,
"frequency": 3,
"token": " Choice",
"token_id": 46785
}
] | {
"arb_Arab": 0.3333333333333333,
"cmn_Hani": null,
"deu_Latn": 0.3333333333333333,
"fra_Latn": 0.3333333333333333
} | {
"arb_Arab": 0.3333333333333333,
"cmn_Hani": null,
"deu_Latn": 0.3333333333333333,
"fra_Latn": 0.3333333333333333
} | feature_10004_complete.json | |
0 | 10,005 | Unknown | No explanation generated | [
" c'est clair. un jeu sans prise de tête pour <<fighter>> les copains"
] | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
14.328125,
0,
0
],
"max_val": 14.328125,
"tokens": [
430,
226,
11,
298,
12231,
18,
270,
3528,
... | 14.328125 | [
{
"average_activation": 14.328125,
"frequency": 1,
"token": " fighter",
"token_id": 69179
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 1
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 1
} | feature_10005_complete.json | |
0 | 10,006 | Unknown | No explanation generated | [
"ordre, flo <<odeurs>> d'IRC, squatteurs de chatrooms",
" am Nuk <<learkurs>> halte man ohnehin fest. \"Ich habe auch keine",
"<<useurs>> ont décidé d'effectuer un nouveau test, et ce, dans",
" and is the natural dividing point between the Ch <<aleurs>> Bay and the",
" - reiner Alkohol zur Reinigung - R... | [
{
"activations": [
0,
0,
0,
0,
0,
11.4296875,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 11.4296875,
"tokens": [
11,
7022,
16,
20142,
2486,
524,
211,
11,
... | 10.593257 | [
{
"average_activation": 10.632900280898877,
"frequency": 89,
"token": "urs",
"token_id": 524
}
] | {
"arb_Arab": null,
"cmn_Hani": 0.010526315789473684,
"deu_Latn": 0.22105263157894736,
"fra_Latn": 0.7684210526315789
} | {
"arb_Arab": null,
"cmn_Hani": 0.010526315789473684,
"deu_Latn": 0.22105263157894736,
"fra_Latn": 0.7684210526315789
} | feature_10006_complete.json | |
0 | 10,007 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10007_complete.json | |
0 | 10,008 | Unknown | No explanation generated | [
" besten Rennspielen für iOS und Android <<,>> was die Fahrphysik und Grafik"
] | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
8.9375,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 8.9375,
"tokens": [
634,
8701,
24946,
36496,
741,
11426,
393,
6975,
... | 8.9375 | [
{
"average_activation": 8.9375,
"frequency": 1,
"token": ",",
"token_id": 16
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 1,
"fra_Latn": null
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 1,
"fra_Latn": null
} | feature_10008_complete.json | |
0 | 10,009 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10009_complete.json | |
0 | 1,000 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_1000_complete.json | |
0 | 10,010 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10010_complete.json | |
0 | 10,011 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10011_complete.json | |
0 | 10,012 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10012_complete.json | |
0 | 10,013 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10013_complete.json | |
0 | 10,014 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10014_complete.json | |
0 | 10,015 | Unknown | No explanation generated | [
"<<oke>> London) permet par exemple à un boulanger de twitter sur son compte une",
" التطوير Ke <<okeN>> Interactive أنّ جهاز مايكروسوفت الاقتصادي للجيل الجديد Xbox Series",
"E. — The Consolidator. The St <<oke>> Newington Daniel Def",
"Cantuccini, Schneemann-Cookies oder Sch <<okokeks>>"
] | [
{
"activations": [
0,
12.4765625,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 12.4765625,
"tokens": [
329,
8981,
8713,
13,
2403,
355,
3552,
321,
... | 11.902344 | [
{
"average_activation": 11.90234375,
"frequency": 4,
"token": "oke",
"token_id": 8981
}
] | {
"arb_Arab": 0.25,
"cmn_Hani": null,
"deu_Latn": 0.25,
"fra_Latn": 0.5
} | {
"arb_Arab": 0.25,
"cmn_Hani": null,
"deu_Latn": 0.25,
"fra_Latn": 0.5
} | feature_10015_complete.json | |
0 | 10,016 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10016_complete.json | |
0 | 10,017 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10017_complete.json | |
0 | 10,018 | Unknown | No explanation generated | [
"chip VitaCell <<之>> 使用方式非常簡單方.. $450.0",
"awa's words 目前分類：2010盛夏 俄羅斯 <<之>> 意想不到之旅 (",
" 2012.02.29 今年橫濱的大雪 <<之>> 日，我們的目的地",
"com同意，不得转载本网站 <<之>> 所有招聘信息及作品；无忧工作网",
" Windows 7 包裝 <<之>> 簡單，也令人驚訝。 用 Windows",
" b. 需更換或退款 <<之>> 產品必須連同未經啟",
"著 Lumia 950 XL <<之>> 圖，但我們相信很快地頭像",
"是由經國家考試合格 <<之>> 消防設備師士、甲種電... | [
{
"activations": [
0,
0,
0,
0,
0,
9.1875,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 9.1875,
"tokens": [
212,
74451,
44650,
77557,
175,
1564,
1902,
7373,
... | 9.102539 | [
{
"average_activation": 9.1025390625,
"frequency": 8,
"token": "之",
"token_id": 1564
}
] | {
"arb_Arab": null,
"cmn_Hani": 1,
"deu_Latn": null,
"fra_Latn": null
} | {
"arb_Arab": null,
"cmn_Hani": 1,
"deu_Latn": null,
"fra_Latn": null
} | feature_10018_complete.json | |
0 | 10,019 | Unknown | No explanation generated | [
" Baumwolle. Alle Drireleas-Gewebe sind mit Fresh <<guard>> ausgestattet"
] | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
13.1015625,
0
],
"max_val": 13.1015625,
"tokens": [
775,
67963,
18,
6176,
3851,
382,
242,
261,
... | 13.101563 | [
{
"average_activation": 13.1015625,
"frequency": 1,
"token": "guard",
"token_id": 24985
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 1,
"fra_Latn": null
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 1,
"fra_Latn": null
} | feature_10019_complete.json | |
0 | 1,001 | Unknown | No explanation generated | [
" des acquis <<techniques>> et théoriques. Cette progression qui peut être suivie au fil des",
" learnt the <<techniques>> that sell across the world, the ways in, which markets",
" les compétences <<techniques,>> la question de la mise en œuvre opérationnelle de cet Observ",
" de nouvelles <<techniques>> ... | [
{
"activations": [
0,
0,
0,
11.1015625,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 11.1015625,
"tokens": [
13813,
314,
11616,
5389,
293,
55644,
18,
35... | 9.737671 | [
{
"average_activation": 10.375679347826088,
"frequency": 69,
"token": " techniques",
"token_id": 5389
},
{
"average_activation": 9.212197580645162,
"frequency": 31,
"token": " technique",
"token_id": 5865
}
] | {
"arb_Arab": 0.01,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 0.99
} | {
"arb_Arab": 0.0078125,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 0.9921875
} | feature_1001_complete.json | |
0 | 10,020 | Unknown | No explanation generated | [
"<<Hybrid-Nanoobjekte>> sind kleinste Verbindungen in der Größe eines ein-mill",
" Nanostrukturen generieren. Die Eigenschaften der Oberfläche der <<Hybrid-Nanoobjekte>>",
" werden neue Strukturen zum Aufbau defin <<ierter Hybrid-Nanoobjekte>> kombiniert und getestet",
"lavendel, der auch als Lavandin bekan... | [
{
"activations": [
0,
12.34375,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 12.34375,
"tokens": [
18,
44482,
17,
50,
8227,
118807,
1294,
88115,
... | 11.878906 | [
{
"average_activation": 12.174479166666666,
"frequency": 3,
"token": " Hybrid",
"token_id": 44482
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 1,
"fra_Latn": null
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 1,
"fra_Latn": null
} | feature_10020_complete.json | |
0 | 10,021 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10021_complete.json | |
0 | 10,022 | Unknown | No explanation generated | [
"C <<rack>> spread)判斷原油價格走勢 雖然說油價與"
] | [
{
"activations": [
0,
0,
14.875,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 14.875,
"tokens": [
12,
39,
51737,
12923,
13,
113679,
59691,
29705,
... | 14.875 | [
{
"average_activation": 14.875,
"frequency": 1,
"token": "rack",
"token_id": 51737
}
] | {
"arb_Arab": null,
"cmn_Hani": 1,
"deu_Latn": null,
"fra_Latn": null
} | {
"arb_Arab": null,
"cmn_Hani": 1,
"deu_Latn": null,
"fra_Latn": null
} | feature_10022_complete.json | |
0 | 10,023 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10023_complete.json | |
0 | 10,024 | Unknown | No explanation generated | [
"<<نا بأن>> الوحدة مستمرة بتقديم خدماتها وفي السياق أثنى السيد مساعد رئيس الجامعة للشؤون",
" مرة. لا تشعر أبدا <<بأنك>> يجب أن تجلس في موقف اللوتس",
"<<فيد بأن>> عملية التصويت في الممثليات الخارجية والمعابر الحدودية للانتخابات البرلمانية والر",
"<<ائل بأنّ>> \"العالم بحاجة إلى العلم، والعلم بحاجة إلى المر... | [
{
"activations": [
0,
0,
11.6484375,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 11.6484375,
"tokens": [
3821,
1441,
4934,
27494,
39313,
26921,
80411,
... | 11.118197 | [
{
"average_activation": 11.4496875,
"frequency": 100,
"token": " بأن",
"token_id": 4934
}
] | {
"arb_Arab": 1,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": null
} | {
"arb_Arab": 0.9455782312925171,
"cmn_Hani": 0.023809523809523808,
"deu_Latn": null,
"fra_Latn": 0.030612244897959183
} | feature_10024_complete.json | |
0 | 10,025 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10025_complete.json | |
0 | 10,026 | Unknown | No explanation generated | [
" Wolken auf. Am Start waren 72 <<Viererbesatzungen,>> darunter 7",
" und Ant <<arktis. Vier>> Jahre Produktionszeit, rund 2.300 Dre",
" in den <<Vierfüsslerstand.>> Ich glaube, dass er eigentlich schon",
" rund 600 auf die Weihnachts <<ferien. Vier>> von fünf Urlaubsreisen an Weihnachten seien",
"schla... | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
19.25,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 19.25,
"tokens": [
94630,
66739,
708,
18,
2556,
8045,
6023,
19477,
... | 15.741175 | [
{
"average_activation": 12.857700892857142,
"frequency": 28,
"token": " بقيادة",
"token_id": 32890
},
{
"average_activation": 18.903645833333332,
"frequency": 24,
"token": " Vier",
"token_id": 15145
}
] | {
"arb_Arab": 0.5185185185185185,
"cmn_Hani": 0.07407407407407407,
"deu_Latn": 0.3888888888888889,
"fra_Latn": 0.018518518518518517
} | {
"arb_Arab": 0.5185185185185185,
"cmn_Hani": 0.07407407407407407,
"deu_Latn": 0.3888888888888889,
"fra_Latn": 0.018518518518518517
} | feature_10026_complete.json | |
0 | 10,027 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10027_complete.json | |
0 | 10,028 | Unknown | No explanation generated | [
"<<tar>> .gz各式的）。 2.直接rpm -ivh ",
". <<tar>> .gz 4. 使用 WP Database Backup 插件把原来的数据库",
"类推。 解压 <<tar>> 打包文件 先cd进入到改压缩文件",
" vor den Tra <<ualtar.>> Das Paar betrieb anschließend gemeinsam ein Gemischtwaren",
": tar -cf fairy <<fish.tar.gz>> * 2. 在新",
"<<blog.wpjam.com/fairyfish.tar.gz>> 3",
"，然后使用 tar... | [
{
"activations": [
0,
16.015625,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 16.015625,
"tokens": [
108062,
44309,
18,
116601,
2776,
26528,
16895,
... | 14.789063 | [
{
"average_activation": 15.513020833333334,
"frequency": 9,
"token": "tar",
"token_id": 44309
}
] | {
"arb_Arab": 0.09090909090909091,
"cmn_Hani": 0.7272727272727273,
"deu_Latn": 0.18181818181818182,
"fra_Latn": null
} | {
"arb_Arab": 0.09090909090909091,
"cmn_Hani": 0.7272727272727273,
"deu_Latn": 0.18181818181818182,
"fra_Latn": null
} | feature_10028_complete.json | |
0 | 10,029 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10029_complete.json | |
0 | 1,002 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_1002_complete.json | |
0 | 10,030 | Unknown | No explanation generated | [
"而成效显著。这款哑光发蜡适合各种长短的头发，",
"大方，戴在手上闪闪亮亮的很是好看，不管是送人",
"... 和中央保持一致 中职(中专)毕业学历人员今年",
"，第一(展现量)第二层(点击量)属于百度推广",
"，做法容易，并附有大厨做汤的秘诀心法",
"反正提议，提示暴躁次倾泻这名单上帝，手虑",
"<<简单>> ，太多会显得杂乱，淹没空间本身的价值，而且",
"。Newcrest Mining大涨3.8%至逾七周",
"韧甲面，呈献极致持久的美甲妆效。同时高",
"”实践活动的延续。 据介绍，2016年3月，镇雄",
"以及可靠性高等一些特点，因此，它可以作为一些大功率模块的... | [
{
"activations": [
9.140625,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 9.140625,
"tokens": [
11368,
52924,
4004,
56171,
424,
19204,
52613,
2... | 8.82581 | [
{
"average_activation": 8.713541666666666,
"frequency": 3,
"token": "简单",
"token_id": 11368
}
] | {
"arb_Arab": null,
"cmn_Hani": 1,
"deu_Latn": null,
"fra_Latn": null
} | {
"arb_Arab": null,
"cmn_Hani": 1,
"deu_Latn": null,
"fra_Latn": null
} | feature_10030_complete.json | |
0 | 10,031 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10031_complete.json | |
0 | 10,032 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10032_complete.json | |
0 | 10,033 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10033_complete.json | |
0 | 10,034 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10034_complete.json | |
0 | 10,035 | Unknown | No explanation generated | [
" appareil photo numérique <<PureView>> avec optique Carl Zeiss grand angle de 41M",
" Mousse Nettoyante Vis <<age Pure>> Clarté. Elle conviendrait aux peaux normales"
] | [
{
"activations": [
0,
0,
0,
0,
12.390625,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 12.390625,
"tokens": [
239,
17498,
5781,
9957,
38445,
24008,
633,
31... | 12.109375 | [
{
"average_activation": 12.109375,
"frequency": 2,
"token": " Pure",
"token_id": 38445
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 1
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 1
} | feature_10035_complete.json | |
0 | 10,036 | Unknown | No explanation generated | [
"’investigation multiprimé, notamment au Dallas <<Morning>> News, Joshua Benton",
" Boston Globe ou le Dallas <<Morning>> News, sont dans une situation différente. Ils",
" FWW GmbH <<, Morningstar>> Deutschland GmbH und weitere. Alle Börsendaten werden",
"'m Wide Awake It <<'s Morning>> Joseph : Ce qu'il ... | [
{
"activations": [
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
13.3984375,
0,
0,
0,
0,
0
],
"max_val": 13.3984375,
"tokens": [
211,
271,
70060,
5125,
7576,
222,
16,
3221,... | 12.84933 | [
{
"average_activation": 13.134114583333334,
"frequency": 6,
"token": " Morning",
"token_id": 49681
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 0.14285714285714285,
"fra_Latn": 0.8571428571428571
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 0.14285714285714285,
"fra_Latn": 0.8571428571428571
} | feature_10036_complete.json | |
0 | 10,037 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10037_complete.json | |
0 | 10,038 | Unknown | No explanation generated | [
" designer de son état, aka @ff_ff sur twitter. Il a"
] | [
{
"activations": [
9.84375,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 9.84375,
"tokens": [
32711,
19980,
236,
697,
8341,
16,
61968,
11160,
... | 9.84375 | [] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 1
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": null,
"fra_Latn": 1
} | feature_10038_complete.json | |
0 | 10,039 | Unknown | No explanation generated | [
" déclin physique et la mort <<tracent>> un portrait amer de la vieillesse. L",
" expérience positive et sans <<tracas.>> Cette équipe est déterminée à vous offrir la",
"er, gr <<ue, tracteurs,>> camions bennes ), un grand déploiement",
" homme <<tracte>> un poids lourd de 15 tonnes à la force de ses musc... | [
{
"activations": [
0,
0,
0,
0,
0,
0,
12,
0,
0,
0,
0,
0,
0,
0,
0,
0
],
"max_val": 12,
"tokens": [
289,
28223,
8780,
293,
279,
3690,
618,
5273,
270,
... | 11.243408 | [
{
"average_activation": 11.490162037037036,
"frequency": 27,
"token": " tra",
"token_id": 618
}
] | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 0.3125,
"fra_Latn": 0.6875
} | {
"arb_Arab": null,
"cmn_Hani": null,
"deu_Latn": 0.3125,
"fra_Latn": 0.6875
} | feature_10039_complete.json | |
0 | 1,003 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_1003_complete.json | |
0 | 10,040 | Unknown | No explanation generated | [] | [] | 0 | [] | null | null | feature_10040_complete.json |
End of preview. Expand in Data Studio
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
GPT2 Multilingual 20% AutoInterp Features
This dataset contains feature interpretations for GPT2 Multilingual model with 20% sparsity.
Structure
data/layer0.parquet- Features for layer 0data/layer1.parquet- Features for layer 1- ...
data/layer11.parquet- Features for layer 11
Each parquet file contains feature interpretation data including:
- Feature activations
- Top examples
- Interpretations
- And other feature analysis data
Usage
from datasets import load_dataset
# Load specific layer
dataset = load_dataset("fdraye/autointerp-gpt2-multilingual-20", data_files="data/layer0.parquet")
# Load all layers
dataset = load_dataset("fdraye/autointerp-gpt2-multilingual-20", data_files="data/*.parquet")
- Downloads last month
- 21