{ "id": "u6r-amp-mid-a0008-g12b", "title": "audit-03 \u00b7 amp typo @ MEASURED mid [28, 31, 34, 37] (\u03b1=0.0075) \u00b7 gemma-12b", "unit": "audit", "created": "2026-07-21T18:35:52", "model": { "name": "gemma-12b", "hf_id": "google/gemma-3-12b-it", "quant": "8bit", "n_layers": 48 }, "lens": { "repo": "neuronpedia/jacobian-lens", "file": "gemma-3-12b-it/jlens/Salesforce-wikitext/gemma-3-12b-it_jacobian_lens.pt" }, "params": { "chat": null, "max_new": 60, "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, 31, 34, 37 ], "mode": "amplify", "alpha": 0.0075 }, "template_kwargs": null, "film": null, "film_start": null, "max_seq_len": null, "lens_layers": null, "temperature": null, "seed": null }, "extra_md": null, "conversation": [ { "role": "user", "content": "Describe the water cycle in two sentences." }, { "role": "assistant", "content": "The water cycle is the continuous process where water evaporates from the earth, rises as water vapour, and then condenses to form clouds. These clouds then release the water back to earth as rain, snow, or other forms of precipitation, which then flows back into bodies of water to start the cycle again" } ], "generated": [ "The water cycle is the continuous process where water evaporates from the earth, rises as water vapour, and then condenses to form clouds. These clouds then release the water back to earth as rain, snow, or other forms of precipitation, which then flows back into bodies of water to start the cycle again" ], "tokens": [ "", "", "user", "\n", "Describe", " the", " water", " cycle", " in", " two", " sentences", ".", "", "\n", "", "model", "\n", "The", " water", " cycle", " is", " the", " continuous", " process", " where", " water", " evaporates", " from", " the", " earth", ",", " rises", " as", " water", " vapour", ",", " and", " then", " cond", "enses", " to", " form", " clouds", ".", " These", " clouds", " then", " release", " the", " water", " back", " to", " earth", " as", " rain", ",", " snow", ",", " or", " other", " forms", " of", " precipitation", ",", " which", " then", " flows", " back", " into", " bodies", " of", " water", " to", " start", " the", " cycle", " again", "", "\n" ], "readouts": [ { "position": 77, "token": "", "model_top": [ ".", "\n", " more", "!", ",", ":", "..", "\n\n" ], "layers": { "0": [ "", "\"\u2014", " \ud83d\udc9e", "", "\ud809\udc2a", " \"\\\\\"", "\ud808\udd4c", " \u7ee7\u7eed\u8bbf\u95ee" ], "1": [ "", "\"\u2014", "\ud809\udc2a", " /**@", "\ud808\udd4c", " \ud83d\udc9e", "\"^", "\uf0be" ], "2": [ "", "\"\u2014", " \ud83d\udc9e", "\ud809\udc2a", "\ud808\udd4c", " /**@", "\u207b", "\ud83c\udd77" ], "3": [ "", " \ud83d\udc9e", "\ud809\udc2a", "\ud808\udd4c", " /**@", "\"\u2014", " \ud83c\udf31", "\u207b" ], "4": [ "", " \ud83d\udc9e", "\"\u2014", "\ud809\udc2a", " \ud83c\udf31", " ```", "\ud808\udd4c", " \"\\\\\"" ], "5": [ "", "\u301d", "\"\u2014", " \ud83d\udc9e", "\uf0be", " /**@", " \ud83c\udf31", "\u06cc" ], "6": [ "", "\"\u2014", "\uf0be", "\u301d", "\u06cc", " /**@", "\u037e", " \ud83d\udc9e" ], "7": [ "", "\"\u2014", "\u06cc", "\uf0be", " /**@", "\u301d", " \ud83d\udc9e", "\ud808\udd4c" ], "8": [ "", "\"\u2014", "\u06cc", " \ud83d\udc9e", "\u301d", "\uf0be", " /**@", "\ud808\udd4c" ], "9": [ "", "\"\u2014", "\u06cc", "\u301d", " \ud83d\udc9e", " /**@", "\uf0be", "\\&" ], "10": [ "", "\u301d", "\u06cc", " \ud83d\udc9e", "\"\u2014", " /**@", " \ud83c\udf31", " \"\"\"\"" ], "11": [ "", " \ud83d\udc9e", "\u301d", "\u06cc", " \ud83d\ude04", " \ud83c\udf31", " /**@", "\u037e" ], "12": [ "", " \ud83d\udc9e", "\u301d", "\u06cc", " \"\"\"\"", " \ud83d\ude04", "\"\"\"\"", " \ud83c\udf31" ], "13": [ "", "\"\u2014", " \ud83d\udc9e", "\u06cc", " \ud83d\ude04", "e", "\\&", "\"^" ], "14": [ "", "\"\u2014", " \ud83d\udc9e", " \ud83d\ude04", "\u06cc", "\\&", "\ud83d\ude04", "e" ], "15": [ "", " \ud83d\ude04", "e", "\u06cc", "\\&", "\ud83d\ude0a", "\ud83d\ude04", "\ud83d\udc4d" ], "16": [ "", " \ud83d\ude04", "e", "\u06cc", "\ud83d\ude04", "\ud83d\ude0a", "\ud83d\udc4d", "Thats" ], "17": [ "", " \ud83d\ude04", "\ud83d\udc49", "\ud83d\ude0a", " \ud83d\udc4d", " \ud83d\udc9e", "\ud83d\udc4d", "\u06cc" ], "18": [ "", "", "\">", " \u7ee7\u7eed\u8bbf\u95ee", " \ud83d\udc4d", " \ud83c\udf31" ], "19": [ "", "", " \ud83d\ude04", "\">", "\ud83d\ude0a", " \">" ], "20": [ "", "", "\">", "\ud83d\ude0a", "\">&" ], "21": [ "", "", "\u17d5", "\">", "", "\">", "\u17d5", "", "\ud83d\udc4d", "\ud83d\ude0a", "\">", "\u17d5", "\ud83d\ude0a", "\ud83d\udc4d", "LinkedIn", "\"/>.", "\ud83c\udf3b", "\">", "", "\u17d5", "\ud83d\ude0a", "README", "\">" ], "26": [ "", "", "\u17d5", "\ud83d\ude0a", "!", "", "\u17d5", "README", "\ud83d\ude0a", "\ud83d\udc4d", "", "" ], "28": [ "", "", "Thats", "\u17d5", " Thats", "!.", " thats", "\ud83d\ude0a" ], "29": [ "", "Thats", " thats", " Thats", "", " whats", " wasnt", "Maybe" ], "30": [ "", "Thats", " thats", " Thats", " whats", " wasnt", " atleast", "" ], "31": [ "", " thats", "!.", " Luckily", " luckily", " Thats", " atleast", " replied" ], "32": [ "", " Luckily", " luckily", "!.", " atleast", " Hopefully", " realises", "Luckily" ], "33": [ "", " Luckily", "!.", "Luckily", " luckily", " Hopefully", " lovely", " Thankfully" ], "34": [ " Luckily", "!.", " luckily", "", "Luckily", " thats", " atleast", " whats" ], "35": [ " Luckily", "!.", " luckily", "Luckily", "", " thats", " whats", " kinda" ], "36": [ " Luckily", " luckily", " piu", " atleast", " alot", "Luckily", "", "!." ], "37": [ " Luckily", " luckily", "!.", " kinda", "Luckily", " hopefully", " word", " Hopefully" ], "38": [ " Luckily", "!.", " luckily", " hopefully", " Hopefully", " Firstly", " kinda", " word" ], "39": [ " Luckily", " luckily", "!.", " hopefully", " Hopefully", " word", " kinda", " nice" ], "40": [ " Luckily", "!.", " hopefully", " luckily", "\n", "", " atleast", " nicely" ], "41": [ "\n", "!.", ".", "", " basically", " more", " hopefully", " Luckily" ], "42": [ ".", "\n", "!.", "!", " .", " more", "..", " with" ], "43": [ "\n", ".", " .", "!.", " more", "!", "..", "" ], "44": [ ".", "\n", " .", "!", "!.", "..", " more", "" ], "45": [ "\n", ".", " .", " more", "!", ":", " ,", ".." ], "46": [ ".", "\n", "!", " more", ",", " .", "..", ":" ] } } ], "trajectories": [ { "word": "anyways", "position": 77, "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 141171, 139376, 141557, 139994, 112680, 143147, 127292, 112151, 62087, 48473, 32172, 24011, 14952, 13488, 10231, 8110, 3586, 5072, 5786, 4011, 4689, 8013, 8204, 14900, 3921, 5634, 15330, 14395, 166, 70, 141, 68, 76, 121, 26, 27, 23, 15, 31, 21, 29, 43, 198, 141, 142, 398, 600 ] }, { "word": "alot", "position": 77, "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 261607, 261391, 259893, 259925, 258426, 260861, 258275, 257464, 255744, 229189, 210607, 132763, 57906, 51954, 54402, 25490, 4389, 51527, 48811, 64937, 41247, 54226, 98754, 139620, 139362, 58482, 164696, 149969, 37, 12, 10, 12, 12, 178, 23, 15, 5, 18, 18, 42, 60, 164, 352, 223, 358, 665, 546 ] }, { "word": "yummy", "position": 77, "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 177869, 180290, 162519, 156509, 141869, 176689, 183931, 191333, 188828, 178563, 159948, 81634, 85164, 100277, 77871, 57074, 110546, 102302, 82527, 64225, 75854, 51727, 43280, 39712, 29263, 25320, 39278, 28614, 407, 231, 252, 49, 39, 47, 88, 78, 51, 58, 85, 107, 212, 1113, 837, 1547, 1763, 3662, 3366 ] }, { "word": "kinda", "position": 77, "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 201830, 192791, 175796, 169481, 144655, 183730, 195226, 186589, 178892, 153703, 128288, 46032, 23136, 26184, 24450, 14085, 7162, 29923, 21117, 13256, 16830, 23722, 31860, 80588, 80548, 54168, 107862, 128559, 346, 101, 84, 29, 46, 71, 20, 7, 12, 4, 7, 7, 13, 34, 115, 200, 336, 256, 221 ] }, { "word": "whilst", "position": 77, "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 136306, 104582, 122046, 141329, 128932, 143922, 124827, 112487, 112752, 52447, 38942, 65485, 18063, 1806, 1018, 1043, 122, 1046, 811, 2007, 2609, 13937, 31849, 122590, 64334, 38554, 57356, 52614, 128, 40, 505, 146, 125, 220, 45, 51, 41, 16, 13, 25, 47, 17, 60, 81, 163, 157, 220 ] }, { "word": "luckily", "position": 77, "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 86671, 96773, 91519, 95558, 86064, 117918, 110587, 113022, 82719, 63998, 45806, 26739, 16211, 8540, 5072, 3584, 4812, 3042, 3396, 3273, 3654, 4763, 3580, 5561, 769, 560, 2759, 3114, 201, 59, 20, 4, 2, 2, 1, 1, 1, 1, 1, 1, 1, 8, 11, 13, 13, 41, 57 ] } ], "emergence": { "position": 77, "top1": ".", "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, 39, 40, 41, 42, 43, 44, 45, 46 ], "ranks": [ 240442, 244536, 232519, 238967, 251568, 259843, 249722, 248221, 239829, 249847, 259632, 249344, 220955, 56424, 18222, 3267, 5605, 10911, 6127, 7529, 2805, 16051, 33475, 178999, 191016, 154048, 178021, 68648, 18, 350, 1525, 64, 150, 107, 95, 52, 101, 44, 22, 19, 11, 3, 1, 2, 1, 2, 1 ] }, "scan": [], "slice": null, "film": null }