| """GPT-2 SAE last-token programs, layer 7 -- features 08000 to 08099 (100 programs). |
| |
| One function per SAE feature. Grouped 100 to a file so the set stays browsable; |
| the programs themselves are unchanged. |
| """ |
|
|
| def prog_L07F08000(input_tokens): |
| """46 token families, 62 surface forms (train F1 of the uncapped optimum: 0.571). |
| Fires at 18,885 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to bravery and courage""" |
| |
| fire = { |
| |
| ' bold', ' boldly', ' Bold', 'bold', |
| |
| ' brave', ' Brave', 'Brave', |
| |
| ' courage', ' courageous', ' Courage', |
| |
| ' heroes', ' heroic', ' Heroic', |
| |
| ' fierce', ' fiercely', |
| |
| ' lonely', ' Lonely', |
| |
| ' warrior', ' warriors', |
| |
| ' pioneer', ' pioneers', |
| |
| ' glorious', |
| |
| ' pioneering', |
| |
| ' patriotic', ' patriot', |
| |
| ' noble', |
| |
| ' daring', |
| |
| ' defiant', |
| |
| ' foolish', |
| |
| ' Feinstein', |
| |
| ' brazen', |
| |
| ' inspiring', |
| |
| ' voluntarily', |
| |
| ' exploits', |
| |
| ' harrowing', |
| |
| ' refreshing', |
| |
| ' Claire', |
| |
| ' willingly', ' willful', |
| |
| ' stubborn', |
| |
| ' adventurous', ' adventurer', |
| |
| 'acity', |
| |
| ' martyr', |
| |
| ' guts', |
| |
| ' blessed', |
| |
| ' comrades', |
| |
| 'ierce', |
| |
| ' bravery', |
| |
| 'roud', |
| |
| ' underdog', |
| |
| ' defiance', |
| |
| ' fearless', |
| |
| ' resilient', |
| |
| ' backbone', |
| |
| ' painstaking', |
| |
| ' spirited', |
| |
| ' admirable', |
| |
| ' disobedience', |
| |
| ' ventured', |
| |
| ' futile', |
| |
| ' traitor', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'bold-family', 1 |
|
|
| def prog_L07F08001(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.072). |
| Fires at 31,369 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating movement or proximity towards something""" |
| |
| fire = { |
| |
| ' to', ' TO', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'to-family', 1 |
|
|
| def prog_L07F08002(input_tokens): |
| """2 token families, 4 surface forms (train F1 of the uncapped optimum: 0.612). |
| Fires at 32,953 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "when" and its context""" |
| |
| fire = { |
| |
| ' when', ' When', ' WHEN', |
| |
| ' whenever', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'when-family', 1 |
|
|
| def prog_L07F08003(input_tokens): |
| """4 token families, 9 surface forms (train F1 of the uncapped optimum: 0.662). |
| Fires at 10,912 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "soon" indicating future events or actions""" |
| |
| fire = { |
| |
| ' soon', ' Soon', 'Soon', 'soon', |
| |
| ' shortly', 'Shortly', ' Shortly', |
| |
| 'Incre', |
| |
| ' ASAP', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'soon-family', 1 |
|
|
| def prog_L07F08004(input_tokens): |
| """47 token families, 53 surface forms (train F1 of the uncapped optimum: 0.562). |
| Fires at 11,783 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the "Oval Office" """ |
| |
| fire = { |
| |
| ' op', |
| |
| ' Oval', ' oval', |
| |
| ' lawn', ' Lawn', |
| |
| ' Cabinet', |
| |
| ' Flip', |
| |
| ' Box', |
| |
| ' recess', |
| |
| ' locker', |
| |
| ' PO', 'PO', |
| |
| ' throne', ' Throne', |
| |
| ' Electoral', |
| |
| ' inbox', |
| |
| ' Downing', |
| |
| ' toilet', |
| |
| ' Keystone', |
| |
| ' lap', |
| |
| ' Ivy', |
| |
| ' pave', |
| |
| ' Roosevelt', |
| |
| 'Lago', |
| |
| ' Flake', |
| |
| ' Pri', |
| |
| ' Neh', |
| |
| ' dispens', |
| |
| ' diary', ' Diary', |
| |
| ' Hyde', |
| |
| ' pent', |
| |
| ' entitle', |
| |
| ' sofa', |
| |
| ' Loft', |
| |
| ' Ely', |
| |
| ' OD', |
| |
| ' Founding', |
| |
| ' Overse', ' overse', |
| |
| ' Matth', |
| |
| ' mailbox', |
| |
| ' Quad', |
| |
| ' corridors', |
| |
| ' enrich', |
| |
| ' medd', |
| |
| 'minster', |
| |
| ' Watergate', |
| |
| ' OB', |
| |
| ' Fountain', |
| |
| ' Malt', |
| |
| ' Abby', |
| |
| ' Prophe', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'op-family', 1 |
|
|
| def prog_L07F08005(input_tokens): |
| """9 token families, 19 surface forms (train F1 of the uncapped optimum: 0.077). |
| Fires at 45,335 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to competitive match formats or rankings""" |
| |
| fire = { |
| |
| ' than', 'than', ' Than', |
| |
| ' known', 'known', |
| |
| 'selling', 'seller', 'sell', ' seller', ' sellers', ' Seller', |
| |
| ' thing', |
| |
| ' possible', |
| |
| ' friend', ' Friend', |
| |
| 'of', |
| |
| ' practices', ' Practices', |
| |
| ' Buy', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'than-family', 1 |
|
|
| def prog_L07F08006(input_tokens): |
| """26 token families, 39 surface forms (train F1 of the uncapped optimum: 0.537). |
| Fires at 18,276 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to formal requests, particularly those related to information or documents""" |
| |
| fire = { |
| |
| ' request', ' Requ', ' Request', 'Request', 'request', 'requ', 'Requ', ' requ', |
| |
| ' requested', ' requests', ' requesting', |
| |
| ' asked', ' Ask', |
| |
| ' demands', |
| |
| 'Rep', |
| |
| ' suggestion', |
| |
| ' Congressman', ' congressman', |
| |
| ' unre', |
| |
| ' invitation', |
| |
| ' inquiries', ' inquired', |
| |
| ' suggestions', |
| |
| ' solic', |
| |
| ' subpoen', |
| |
| ' forcibly', |
| |
| ' recommendation', |
| |
| ' pleas', |
| |
| ' query', |
| |
| ' subpoena', |
| |
| ' behest', |
| |
| ' solicit', |
| |
| ' FOIA', |
| |
| ' Beg', 'Beg', |
| |
| ' politely', |
| |
| ' queries', |
| |
| ' inappropriately', |
| |
| ' petitions', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'requ-family', 1 |
|
|
| def prog_L07F08007(input_tokens): |
| """36 token families, 46 surface forms (train F1 of the uncapped optimum: 0.706). |
| Fires at 11,641 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of restaurants""" |
| |
| fire = { |
| |
| ' restaurant', ' restaurants', ' Restaurant', |
| |
| ' chef', ' chefs', ' Chef', |
| |
| ' dining', ' din', ' diner', |
| |
| ' eater', ' Eater', |
| |
| 'food', |
| |
| ' vegetarian', |
| |
| ' Tavern', ' tavern', |
| |
| ' Grill', |
| |
| ' cuisine', |
| |
| ' Cul', |
| |
| ' patrons', |
| |
| ' Naj', |
| |
| ' culinary', |
| |
| ' establishments', |
| |
| ' Pub', |
| |
| ' restaur', ' Restaur', |
| |
| ' seafood', |
| |
| ' Gast', ' gast', |
| |
| 'ushi', |
| |
| 'essen', |
| |
| 'unct', |
| |
| ' Chinatown', |
| |
| ' Subway', |
| |
| ' taco', |
| |
| ' Cheng', |
| |
| ' waitress', |
| |
| 'otle', |
| |
| ' Yelp', |
| |
| 'uten', |
| |
| ' hospitality', |
| |
| 'Rest', |
| |
| ' waiter', |
| |
| 'ourmet', |
| |
| ' sushi', |
| |
| ' Seaf', |
| |
| 'ateur', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'restaurant-family', 1 |
|
|
| def prog_L07F08008(input_tokens): |
| """12 token families, 13 surface forms (train F1 of the uncapped optimum: 0.136). |
| Fires at 21,367 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to educational institutions, specifically focusing on the word "school." """ |
| |
| fire = { |
| |
| 'ol', |
| |
| 'IC', |
| |
| 'ab', |
| |
| 'um', |
| |
| 'app', |
| |
| 'uster', |
| |
| 'ulz', |
| |
| 'iller', 'illing', |
| |
| 'ef', |
| |
| 'ä', |
| |
| 'Name', |
| |
| 'ng', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ol-family', 1 |
|
|
| def prog_L07F08009(input_tokens): |
| """20 token families, 51 surface forms (train F1 of the uncapped optimum: 0.118). |
| Fires at 1,207,283 of 36,786,566 positions. |
| Fires at 3.28% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: references to steps in a process or sequential actions""" |
| |
| fire = { |
| |
| ',', ' ,', |
| |
| ' and', ' And', 'and', 'And', ' AND', 'ander', 'ands', 'anded', |
| |
| ' to', |
| |
| ' that', ' That', ' THAT', ' thats', |
| |
| ' (', |
| |
| ' The', |
| |
| ' it', |
| |
| ':', ' :', |
| |
| ')', ' )', |
| |
| ' will', ' WILL', ' willingly', |
| |
| ' —', '—', |
| |
| ' one', ' One', ' ones', 'one', ' ONE', ' Ones', 'ONES', |
| |
| ' he', |
| |
| ' can', 'Can', ' cans', |
| |
| ' or', ' Or', ' OR', |
| |
| ' but', |
| |
| ' said', |
| |
| ' when', ' When', 'when', |
| |
| ' they', ' THEY', |
| |
| ' would', 'Would', ' Would', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L07F08010(input_tokens): |
| """7 token families, 11 surface forms (train F1 of the uncapped optimum: 0.121). |
| Fires at 179,714 of 36,786,566 positions. |
| Fires at 0.49% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: references to America's social issues and challenges""" |
| |
| fire = { |
| |
| 's', "'s", |
| |
| ' its', ' Its', ' ITS', |
| |
| ' our', ' Our', |
| |
| 're', |
| |
| ' should', |
| |
| ' America', |
| |
| ' economy', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 's-family', 1 |
|
|
| def prog_L07F08011(input_tokens): |
| """17 token families, 62 surface forms (train F1 of the uncapped optimum: 0.148). |
| Fires at 330,947 of 36,786,566 positions. |
| Fires at 0.90% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: references to sexual misconduct and exploitation involving minors""" |
| |
| fire = { |
| |
| ' of', ' OF', 'OF', |
| |
| ' with', ' WITH', |
| |
| ' her', ' Her', 'hered', ' HER', ' hers', |
| |
| 'old', ' old', ' older', ' oldest', 'Old', ' Older', 'older', |
| |
| ' their', 'their', ' THEIR', |
| |
| ' young', ' younger', ' youngest', ' Young', 'Young', 'young', |
| |
| ' two', 'two', ' TWO', |
| |
| ' them', ' Them', 'them', |
| |
| ' my', 'my', ' MY', |
| |
| ' him', ' Him', 'him', |
| |
| ' your', ' Your', 'your', ' YOUR', |
| |
| ' girl', ' girls', ' Girl', 'girl', ' Girls', 'girls', 'Girls', 'Girl', |
| |
| ' me', ' Me', |
| |
| ' she', 'she', |
| |
| ' little', ' Little', 'Little', 'little', |
| |
| ' around', 'around', |
| |
| 'year', 'Year', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'of-family', 1 |
|
|
| def prog_L07F08012(input_tokens): |
| """7 token families, 13 surface forms (train F1 of the uncapped optimum: 0.048). |
| Fires at 46,472 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating comparison or additional information""" |
| |
| fire = { |
| |
| ':', |
| |
| ' because', |
| |
| 'Except', ' except', ' Except', 'except', |
| |
| ' thing', |
| |
| ' fact', |
| |
| ' Plus', 'Plus', ' PLUS', |
| |
| ' worse', ' Worse', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ':-family', 1 |
|
|
| def prog_L07F08013(input_tokens): |
| """40 token families, 50 surface forms (train F1 of the uncapped optimum: 0.724). |
| Fires at 24,029 of 36,786,566 positions.""" |
| |
| """prev SAE description: words that denote capability or possibility""" |
| |
| fire = { |
| |
| 'able', 'ABLE', |
| |
| 'ability', 'abil', 'abilities', |
| |
| 'ible', 'IBLE', |
| |
| 'ables', 'abled', 'abl', |
| |
| 'ably', |
| |
| ' playable', 'playing', |
| |
| 'ibles', 'ibl', |
| |
| ' compatible', 'compatible', |
| |
| 'borne', |
| |
| 'worthy', |
| |
| 'itable', |
| |
| 'iable', |
| |
| ' wearable', |
| |
| 'ibility', |
| |
| 'resistant', |
| |
| 'orable', |
| |
| ' usable', 'usable', |
| |
| ' downloadable', |
| |
| 'yle', |
| |
| 'hetic', |
| |
| 'urable', |
| |
| 'atable', |
| |
| 'onomous', |
| |
| 'ifiable', |
| |
| ' durable', |
| |
| ' punishable', |
| |
| ' edible', |
| |
| ' disposable', |
| |
| 'enabled', |
| |
| ' reusable', |
| |
| 'latable', |
| |
| ' scalable', |
| |
| ' customizable', |
| |
| 'achable', |
| |
| 'ovable', |
| |
| 'uable', |
| |
| ' adjustable', |
| |
| ' finite', |
| |
| ' convertible', |
| |
| 'uble', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'able-family', 1 |
|
|
| def prog_L07F08014(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.322). |
| Fires at 157,975 of 36,786,566 positions. |
| Fires at 0.43% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: instances of the word "the." """ |
| |
| fire = { |
| |
| ' The', 'the', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'the-family', 1 |
|
|
| def prog_L07F08015(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.032). |
| Fires at 13,399 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases that begin with "any of" followed by a subset of items or concepts""" |
| |
| fire = { |
| |
| ' of', ' OF', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'of-family', 1 |
|
|
| def prog_L07F08016(input_tokens): |
| """48 token families, 48 surface forms (train F1 of the uncapped optimum: 0.503). |
| Fires at 192,049 of 36,786,566 positions. |
| Fires at 0.52% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: instances of personal pronouns and discussions about individual perspectives or actions""" |
| |
| fire = { |
| |
| 'The', |
| |
| 'It', |
| |
| 'In', |
| |
| 'I', |
| |
| 'This', |
| |
| 'He', |
| |
| 'We', |
| |
| 'That', |
| |
| 'As', |
| |
| 'There', |
| |
| 'According', |
| |
| 'They', |
| |
| 'So', |
| |
| 'On', |
| |
| 'You', |
| |
| 'While', |
| |
| 'Here', |
| |
| 'One', |
| |
| 'For', |
| |
| 'Now', |
| |
| 'She', |
| |
| 'When', |
| |
| 'What', |
| |
| 'After', |
| |
| 'At', |
| |
| 'And', |
| |
| 'With', |
| |
| 'However', |
| |
| 'His', |
| |
| 'Mr', |
| |
| 'These', |
| |
| 'Some', |
| |
| 'My', |
| |
| 'During', |
| |
| 'All', |
| |
| 'From', |
| |
| 'Speaking', |
| |
| 'Not', |
| |
| 'Since', |
| |
| 'Well', |
| |
| 'Many', |
| |
| 'Let', |
| |
| 'Both', |
| |
| 'Her', |
| |
| 'Then', |
| |
| 'Even', |
| |
| 'Those', |
| |
| 'Instead', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'the-family', 1 |
|
|
| def prog_L07F08017(input_tokens): |
| """9 token families, 11 surface forms (train F1 of the uncapped optimum: 0.066). |
| Fires at 52,184 of 36,786,566 positions.""" |
| |
| """prev SAE description: timestamps and dates in the text""" |
| |
| fire = { |
| |
| ':', ' :', |
| |
| ' –', |
| |
| '):', |
| |
| ' 18', |
| |
| ' |', |
| |
| ' 17', |
| |
| ' 23', |
| |
| ' marks', ' marked', |
| |
| ' 24', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ':-family', 1 |
|
|
| def prog_L07F08018(input_tokens): |
| """13 token families, 23 surface forms (train F1 of the uncapped optimum: 0.155). |
| Fires at 199,491 of 36,786,566 positions. |
| Fires at 0.54% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: terms related to legal classifications and offenses""" |
| |
| fire = { |
| |
| ' as', |
| |
| ' like', ' LIKE', |
| |
| ' called', 'called', ' Called', |
| |
| ' really', ' real', ' Really', ' REALLY', ' REAL', |
| |
| " '", |
| |
| ' part', |
| |
| ' become', ' Become', |
| |
| ' considered', |
| |
| ' known', |
| |
| ' true', 'true', |
| |
| ' officially', |
| |
| ' actually', 'actually', |
| |
| ' potential', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'as-family', 1 |
|
|
| def prog_L07F08019(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.030). |
| Fires at 21,583 of 36,786,566 positions.""" |
| |
| """prev SAE description: numerical data or dates associated with events""" |
| |
| fire = { |
| |
| '\n', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '\\n-family', 1 |
|
|
| def prog_L07F08020(input_tokens): |
| """49 token families, 56 surface forms (train F1 of the uncapped optimum: 0.583). |
| Fires at 9,260 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to dial systems and their functionalities""" |
| |
| fire = { |
| |
| ' dial', ' Dial', 'dial', 'Dial', |
| |
| ' Monte', |
| |
| ' Vin', |
| |
| ' Dixon', |
| |
| ' winding', |
| |
| ' caller', ' CALL', |
| |
| ' brass', |
| |
| ' Vo', |
| |
| ' dialog', 'Dialog', |
| |
| ' Vij', |
| |
| ' Tol', |
| |
| ' dialect', |
| |
| ' rou', |
| |
| ' voic', |
| |
| ' Hawking', |
| |
| 'oppers', |
| |
| ' ticking', |
| |
| 'reet', |
| |
| ' telev', |
| |
| ' Tek', |
| |
| ' hooked', |
| |
| ' CPR', |
| |
| ' calculator', ' Calculator', |
| |
| ' hotline', |
| |
| ' countdown', |
| |
| ' PIN', |
| |
| ' Rai', |
| |
| ' rang', |
| |
| ' cooker', |
| |
| 'iven', |
| |
| ' bells', |
| |
| ' cho', |
| |
| 'tone', ' Tone', |
| |
| ' jargon', |
| |
| ' Ware', |
| |
| ' dispatcher', |
| |
| ' radios', |
| |
| ' Vide', |
| |
| ' ration', |
| |
| ' Roche', |
| |
| 'tel', |
| |
| ' Chak', |
| |
| ' Sloan', |
| |
| ' Diver', |
| |
| ' fax', |
| |
| ' intraven', |
| |
| ' Geral', |
| |
| ' Viet', |
| |
| ' cyt', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'dial-family', 1 |
|
|
| def prog_L07F08021(input_tokens): |
| """7 token families, 9 surface forms (train F1 of the uncapped optimum: 0.128). |
| Fires at 371,308 of 36,786,566 positions. |
| Fires at 1.01% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: terms and phrases related to emergency situations and chaos""" |
| |
| fire = { |
| |
| '\n', |
| |
| ' caption', |
| |
| ' —', |
| |
| ',"', |
| |
| 'As', ' As', |
| |
| ' according', 'According', |
| |
| '."', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '\\n-family', 1 |
|
|
| def prog_L07F08022(input_tokens): |
| """10 token families, 16 surface forms (train F1 of the uncapped optimum: 0.112). |
| Fires at 202,317 of 36,786,566 positions. |
| Fires at 0.55% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: prepositions and their use in context""" |
| |
| fire = { |
| |
| ' of', |
| |
| ' for', |
| |
| ' on', |
| |
| ' from', ' FROM', |
| |
| ' about', ' ABOUT', |
| |
| ' between', 'between', |
| |
| ' against', ' Against', 'against', |
| |
| ' during', 'during', |
| |
| ' across', |
| |
| ' within', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'of-family', 1 |
|
|
| def prog_L07F08023(input_tokens): |
| """13 token families, 36 surface forms (train F1 of the uncapped optimum: 0.120). |
| Fires at 203,687 of 36,786,566 positions. |
| Fires at 0.55% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases expressing strong sentiments or deep reflections""" |
| |
| fire = { |
| |
| "'", |
| |
| '"', |
| |
| '!', |
| |
| ' world', ' worlds', ' Worlds', ' WORLD', ' worldly', |
| |
| ' life', ' Life', ' LIFE', |
| |
| ' things', ' thing', ' Things', ' Thing', 'thing', 'things', |
| |
| ' again', ' Again', 'again', |
| |
| ' me', ' Me', ' ME', |
| |
| ' happen', ' happening', ' happened', ' happens', |
| |
| ' here', ' HERE', |
| |
| ' email', |
| |
| ' easy', ' Easy', 'Easy', 'easy', |
| |
| ' trust', ' trusted', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return "'-family", 1 |
|
|
| def prog_L07F08024(input_tokens): |
| """21 token families, 22 surface forms (train F1 of the uncapped optimum: 0.193). |
| Fires at 53,315 of 36,786,566 positions.""" |
| |
| """prev SAE description: graphic descriptions of violence or harm""" |
| |
| fire = { |
| |
| ' death', |
| |
| ' end', |
| |
| ' hospital', |
| |
| ' ground', |
| |
| ' level', |
| |
| ' scene', |
| |
| ' floor', |
| |
| ' bottom', |
| |
| ' streets', |
| |
| ' blame', |
| |
| ' polls', |
| |
| ' wall', |
| |
| ' conclusion', |
| |
| ' truth', |
| |
| ' door', |
| |
| ' table', |
| |
| ' jail', |
| |
| ' ballot', |
| |
| ' moon', ' Moon', |
| |
| ' extent', |
| |
| ' bed', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'death-family', 1 |
|
|
| def prog_L07F08025(input_tokens): |
| """47 token families, 51 surface forms (train F1 of the uncapped optimum: 0.741). |
| Fires at 9,344 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the Czech Republic and its capital, Prague""" |
| |
| fire = { |
| |
| ' Czech', |
| |
| ' Idaho', |
| |
| ' CES', |
| |
| 'jiang', ' Jiang', |
| |
| ' Gerald', |
| |
| 'cz', ' cz', |
| |
| ' Chile', |
| |
| ' Tibet', |
| |
| ' cough', ' coughing', |
| |
| ' Croatia', |
| |
| ' Tibetan', |
| |
| ' Dominican', |
| |
| 'č', |
| |
| ' Baltic', |
| |
| ' Cork', |
| |
| ' Cameroon', |
| |
| ' Hebrew', |
| |
| ' Boyle', |
| |
| ' Jak', |
| |
| ' CBI', |
| |
| 'ague', |
| |
| 'ć', |
| |
| 'Desc', ' desc', |
| |
| ' Buch', |
| |
| ' Dominic', |
| |
| 'š', |
| |
| ' DSL', |
| |
| ' Prague', |
| |
| ' Slovakia', |
| |
| ' Croatian', |
| |
| ' Slav', |
| |
| ' dict', |
| |
| ' Slov', |
| |
| 'oqu', |
| |
| ' Cambodia', |
| |
| ' Budapest', |
| |
| ' Boh', |
| |
| ' Latvia', |
| |
| ' disinteg', |
| |
| ' Siem', |
| |
| ' Thib', |
| |
| 'decl', |
| |
| ' CST', |
| |
| ' Bohem', |
| |
| ' Pavel', |
| |
| ' Sloven', |
| |
| ' drib', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'czech-family', 1 |
|
|
| def prog_L07F08026(input_tokens): |
| """9 token families, 24 surface forms (train F1 of the uncapped optimum: 0.127). |
| Fires at 523,150 of 36,786,566 positions. |
| Fires at 1.42% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases related to international relations and conflict""" |
| |
| fire = { |
| |
| ',', ' ,', |
| |
| ' years', ' year', ' Year', ' YEAR', |
| |
| ' (', |
| |
| ')', |
| |
| '"', |
| |
| ' time', 'time', ' TIME', ' timely', |
| |
| ' state', ' State', 'state', ' statements', ' Statement', 'statement', |
| |
| ' week', ' weeks', |
| |
| ' month', ' months', 'months', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L07F08027(input_tokens): |
| """4 token families, 7 surface forms (train F1 of the uncapped optimum: 0.058). |
| Fires at 54,670 of 36,786,566 positions.""" |
| |
| """prev SAE description: superlative statements and phrases of comparison""" |
| |
| fire = { |
| |
| ' the', 'the', ' THE', |
| |
| ' most', |
| |
| ' arguably', |
| |
| ' perhaps', 'Perhaps', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'the-family', 1 |
|
|
| def prog_L07F08028(input_tokens): |
| """45 token families, 54 surface forms (train F1 of the uncapped optimum: 0.593). |
| Fires at 12,248 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of delegates in a political context""" |
| |
| fire = { |
| |
| ' deputies', |
| |
| ' delegates', ' delegated', |
| |
| ' attendees', |
| |
| ' Electoral', ' electors', ' elector', |
| |
| ' delegation', |
| |
| ' diplomats', ' diplomat', |
| |
| ' councillors', ' councillor', |
| |
| ' negotiators', ' negotiator', |
| |
| ' miners', |
| |
| ' MEP', |
| |
| ' commissioners', ' Commissioners', |
| |
| 'uties', |
| |
| ' Derrick', |
| |
| ' delegate', |
| |
| ' allegiance', |
| |
| ' Whip', ' whip', |
| |
| ' MLA', |
| |
| 'legates', |
| |
| ' creditors', |
| |
| 'anche', |
| |
| ' supremacists', |
| |
| ' Rosa', |
| |
| ' Heidi', |
| |
| ' trustees', |
| |
| ' Lenin', 'Lenin', |
| |
| ' runaway', |
| |
| ' caucuses', |
| |
| ' DUP', |
| |
| ' evangelicals', |
| |
| ' Kem', |
| |
| ' DeVos', |
| |
| ' ratified', |
| |
| ' Witnesses', |
| |
| ' troopers', |
| |
| 'arde', |
| |
| 'legate', |
| |
| ' trustee', |
| |
| ' proportional', |
| |
| ' deleg', |
| |
| ' transistor', |
| |
| ' mitochondrial', |
| |
| ' Connie', |
| |
| ' adherents', |
| |
| 'wit', |
| |
| ' ratification', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'deput-family', 1 |
|
|
| def prog_L07F08029(input_tokens): |
| """3 token families, 4 surface forms (train F1 of the uncapped optimum: 0.125). |
| Fires at 60,907 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to deaths and killings in various contexts""" |
| |
| fire = { |
| |
| ' in', |
| |
| ' during', 'during', |
| |
| ' killed', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'in-family', 1 |
|
|
| def prog_L07F08030(input_tokens): |
| """16 token families, 45 surface forms (train F1 of the uncapped optimum: 0.070). |
| Fires at 502,822 of 36,786,566 positions. |
| Fires at 1.37% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: words indicating legal terminology or concepts""" |
| |
| fire = { |
| |
| ',', ' ,', |
| |
| ' is', ' IS', |
| |
| ' it', ' It', 'ité', |
| |
| ' (', |
| |
| ' are', ' ARE', 'ares', |
| |
| ' you', ' You', ' YOU', |
| |
| ' I', ' i', |
| |
| ' but', ' But', 'BUT', ' BUT', |
| |
| ' we', ' We', ' WE', |
| |
| ' they', ' They', ' THEY', |
| |
| ' more', ' More', |
| |
| ' all', ' All', ' ALL', 'alled', ' Allies', |
| |
| ' just', ' Just', ' JUST', |
| |
| ' so', ' So', ' SO', |
| |
| ' there', ' There', 'thereal', |
| |
| ' what', ' What', ' WHAT', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L07F08031(input_tokens): |
| """5 token families, 12 surface forms (train F1 of the uncapped optimum: 0.826). |
| Fires at 7,509 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to various fields or areas of study""" |
| |
| fire = { |
| |
| ' field', ' fields', ' Field', ' Fields', ' fielded', 'fields', ' fielding', 'Field', |
| |
| ' battlefield', |
| |
| ' constituency', |
| |
| ' countryside', |
| |
| 'irie', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'field-family', 1 |
|
|
| def prog_L07F08032(input_tokens): |
| """14 token families, 29 surface forms (train F1 of the uncapped optimum: 0.781). |
| Fires at 18,823 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "move" and its variations""" |
| |
| fire = { |
| |
| ' moving', ' moved', ' moves', 'moving', ' Moving', 'Moving', ' Moves', ' mov', ' MOV', |
| |
| ' move', ' Move', 'Move', 'move', |
| |
| ' shifted', ' shifting', |
| |
| ' migrating', ' migrated', |
| |
| ' relocate', |
| |
| ' relocated', |
| |
| ' relocation', |
| |
| ' creeping', ' creeps', |
| |
| ' migrate', |
| |
| ' shuff', |
| |
| ' transitioning', ' transitioned', |
| |
| 'ocating', |
| |
| ' adjourn', |
| |
| ' bounced', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'mov-family', 1 |
|
|
| def prog_L07F08033(input_tokens): |
| """22 token families, 26 surface forms (train F1 of the uncapped optimum: 0.173). |
| Fires at 167,059 of 36,786,566 positions. |
| Fires at 0.45% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: words associated with organized activities, publications, and events""" |
| |
| fire = { |
| |
| ' series', |
| |
| ' team', |
| |
| ' police', |
| |
| ' reports', ' reporter', |
| |
| ' program', |
| |
| ' project', |
| |
| ' website', |
| |
| ' event', |
| |
| ' magazine', |
| |
| ' app', |
| |
| ' service', |
| |
| ' party', |
| |
| ' community', |
| |
| ' award', ' awards', ' Award', |
| |
| ' area', |
| |
| ' stories', ' stores', |
| |
| ' page', |
| |
| ' star', |
| |
| ' store', |
| |
| ' poll', |
| |
| ' breaking', |
| |
| ' conference', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ser-family', 1 |
|
|
| def prog_L07F08034(input_tokens): |
| """31 token families, 51 surface forms (train F1 of the uncapped optimum: 0.682). |
| Fires at 86,449 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to centralized control or authority""" |
| |
| fire = { |
| |
| ' unveiled', ' unveil', ' unveiling', |
| |
| ' iconic', |
| |
| ' spokesperson', |
| |
| ' newest', |
| |
| ' outlined', ' outlining', ' outlines', |
| |
| ' focused', 'focused', |
| |
| ' thrilled', ' thriller', ' thrilling', |
| |
| ' flagship', |
| |
| ' deployed', ' deploying', ' deploy', |
| |
| ' touted', ' touting', |
| |
| ' showcase', |
| |
| ' acclaimed', ' acclaim', |
| |
| ' debuted', |
| |
| ' technologies', |
| |
| ' advisor', ' advisors', ' Advisor', |
| |
| ' centered', 'centered', |
| |
| ' unleashed', ' unleash', ' Unleashed', |
| |
| ' marketplace', ' Marketplace', |
| |
| ' cryptocurrency', |
| |
| 'driven', |
| |
| ' arguably', |
| |
| ' superstar', |
| |
| ' comprehensive', |
| |
| ' tasked', |
| |
| ' overview', ' Overview', |
| |
| ' unified', ' Unified', |
| |
| ' cybersecurity', |
| |
| ' foremost', |
| |
| ' overwhelmed', ' overwhelm', |
| |
| ' forefront', |
| |
| ' governance', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'unveil-family', 1 |
|
|
| def prog_L07F08035(input_tokens): |
| """22 token families, 49 surface forms (train F1 of the uncapped optimum: 0.139). |
| Fires at 72,544 of 36,786,566 positions.""" |
| |
| """prev SAE description: words related to applications and their functionality""" |
| |
| fire = { |
| |
| '-', |
| |
| '/', |
| |
| 'a', 'ña', |
| |
| 'al', 'AL', |
| |
| 'en', 'EN', |
| |
| ' app', ' App', 'app', 'App', 'apps', ' Apps', ' APP', |
| |
| 'OL', 'ol', |
| |
| 'o', |
| |
| 'earing', 'ear', 'eared', 'ears', 'early', 'earance', |
| |
| 'le', 'LE', |
| |
| ' Store', 'store', 'Store', |
| |
| 'il', |
| |
| 'l', 'lé', |
| |
| ' Sports', ' Sport', ' SPORTS', |
| |
| 'el', |
| |
| 'et', 'ET', |
| |
| 'man', |
| |
| 'is', |
| |
| 'les', 'less', 'LES', |
| |
| 'os', 'OS', |
| |
| ' Pinterest', |
| |
| 'ION', 'ion', 'ional', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '--family', 1 |
|
|
| def prog_L07F08036(input_tokens): |
| """36 token families, 38 surface forms (train F1 of the uncapped optimum: 0.612). |
| Fires at 8,133 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to legal charges and indictments""" |
| |
| fire = { |
| |
| ' indictment', ' indicted', ' indict', |
| |
| ' impeachment', |
| |
| ' damning', |
| |
| ' Abdel', |
| |
| 'VERTIS', |
| |
| ' extradition', |
| |
| ' biography', |
| |
| ' defendants', |
| |
| ' unse', |
| |
| ' arra', |
| |
| ' autopsy', |
| |
| ' impe', |
| |
| ' Taj', |
| |
| ' unm', |
| |
| ' antitrust', |
| |
| ' electr', |
| |
| ' briefed', |
| |
| ' Sach', |
| |
| ' motherboard', |
| |
| ' extrad', |
| |
| ' conspiring', |
| |
| ' implicated', |
| |
| ' criminally', |
| |
| ' Sergei', |
| |
| ' inquest', |
| |
| ' bullish', |
| |
| 'mortem', |
| |
| 'mys', |
| |
| ' incrim', |
| |
| ' architectures', |
| |
| ' ben', |
| |
| ' culp', |
| |
| ' rive', |
| |
| ' indis', |
| |
| ' boiler', |
| |
| 'achment', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'indict-family', 1 |
|
|
| def prog_L07F08037(input_tokens): |
| """16 token families, 24 surface forms (train F1 of the uncapped optimum: 0.173). |
| Fires at 165,625 of 36,786,566 positions. |
| Fires at 0.45% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: significant nouns related to legal and procedural contexts""" |
| |
| fire = { |
| |
| ' system', ' System', |
| |
| ' game', ' GAME', |
| |
| ' program', |
| |
| ' process', ' procession', |
| |
| ' attack', 'attack', |
| |
| ' bill', |
| |
| ' event', ' Event', |
| |
| ' election', 'election', |
| |
| ' market', |
| |
| ' crisis', |
| |
| ' race', |
| |
| ' rate', |
| |
| ' tax', |
| |
| ' ban', |
| |
| ' award', ' Award', |
| |
| ' conference', 'conference', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'system-family', 1 |
|
|
| def prog_L07F08038(input_tokens): |
| """9 token families, 17 surface forms (train F1 of the uncapped optimum: 0.724). |
| Fires at 11,078 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases that indicate expectations or anticipated outcomes""" |
| |
| fire = { |
| |
| ' expected', ' expect', ' expecting', 'expected', ' Expect', |
| |
| ' anticipated', 'anticipated', ' anticipating', |
| |
| ' unexpected', ' unexpectedly', |
| |
| ' predictable', |
| |
| ' expectancy', |
| |
| 'pected', 'pecting', |
| |
| ' guessed', |
| |
| ' anticipate', |
| |
| 'nexpected', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'expect-family', 1 |
|
|
| def prog_L07F08039(input_tokens): |
| """48 token families, 57 surface forms (train F1 of the uncapped optimum: 0.684). |
| Fires at 19,275 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the name "Molly" or variations of it""" |
| |
| fire = { |
| |
| ' bomber', ' bombers', ' Bomber', |
| |
| 'ONDON', |
| |
| 'olly', |
| |
| 'zy', |
| |
| 'psy', ' Psy', ' psy', 'Psy', |
| |
| ' Holmes', |
| |
| 'oli', |
| |
| 'oly', |
| |
| 'oko', |
| |
| ' Barrett', |
| |
| ' Volkswagen', |
| |
| ' Katie', |
| |
| 'otti', |
| |
| '559', |
| |
| 'oping', |
| |
| 'erry', |
| |
| 'acey', |
| |
| 'OO', |
| |
| ' Oxy', 'oxy', |
| |
| ' Rolling', |
| |
| ' puppy', |
| |
| ' Holly', ' HOL', |
| |
| ' psycho', ' Psycho', |
| |
| 'ooky', |
| |
| ' Neo', |
| |
| 'ppy', |
| |
| ' Marty', |
| |
| 'ihadi', |
| |
| 'oco', |
| |
| ' Bom', |
| |
| ' VW', 'VW', |
| |
| ' chemotherapy', |
| |
| 'oby', |
| |
| ' payload', |
| |
| 'ocol', |
| |
| ' workload', |
| |
| ' luggage', |
| |
| ' Wolfe', |
| |
| 'acco', |
| |
| ' obsessive', |
| |
| ' Molly', |
| |
| 'eddy', |
| |
| 'ortun', |
| |
| ' pillow', |
| |
| ' Woody', |
| |
| ' Nightmare', |
| |
| 'traumatic', |
| |
| ' blo', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'bomb-family', 1 |
|
|
| def prog_L07F08040(input_tokens): |
| """38 token families, 57 surface forms (train F1 of the uncapped optimum: 0.671). |
| Fires at 11,238 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "confusion" and its variants""" |
| |
| fire = { |
| |
| ' conf', ' Conf', 'conf', 'Conf', |
| |
| 'Con', ' CON', 'CON', |
| |
| 'Ill', ' Ill', |
| |
| ' confidential', |
| |
| ' Enc', 'enc', 'Enc', |
| |
| ' Securities', |
| |
| ' Reve', 'reve', |
| |
| 'Ind', |
| |
| ' Obs', 'Obs', |
| |
| 'Inc', 'INC', |
| |
| ' ded', ' Ded', |
| |
| 'Calif', |
| |
| ' reaff', |
| |
| ' Strange', |
| |
| 'Diff', 'diff', ' Diff', |
| |
| ' coinc', |
| |
| ' Coord', 'coord', ' coord', |
| |
| 'Comm', |
| |
| ' Prep', 'prep', |
| |
| 'Tenn', |
| |
| ' confid', |
| |
| 'clinton', |
| |
| 'CN', 'cn', |
| |
| 'Commun', ' COMMUN', |
| |
| 'prof', |
| |
| ' Consult', |
| |
| ' priv', |
| |
| ' encl', |
| |
| ' Spec', |
| |
| ' Columb', |
| |
| ' perplex', |
| |
| ' unn', |
| |
| ' Conc', |
| |
| 'idential', |
| |
| 'CLA', |
| |
| ' confidentiality', |
| |
| ' PRES', |
| |
| ' CLE', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'conf-family', 1 |
|
|
| def prog_L07F08041(input_tokens): |
| """46 token families, 65 surface forms (train F1 of the uncapped optimum: 0.657). |
| Fires at 10,672 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to filesystem paths""" |
| |
| fire = { |
| |
| ' path', ' Path', ' paths', 'path', 'Path', 'pathic', ' PATH', 'PATH', |
| |
| ' Math', 'math', 'Math', |
| |
| ' ES', |
| |
| 'yll', |
| |
| ' Gol', |
| |
| ' Kathy', |
| |
| ' URL', ' URLs', ' url', 'Url', |
| |
| ' Tales', |
| |
| ' Christine', |
| |
| ' pathways', ' pathway', |
| |
| ' Parkway', |
| |
| ' Gateway', |
| |
| 'ATH', |
| |
| ' Pamela', |
| |
| ' Pale', |
| |
| ' UNC', |
| |
| ' Olson', 'olson', |
| |
| ' Crescent', |
| |
| ' prot', |
| |
| 'eson', |
| |
| ' Norwich', |
| |
| ' psychopath', |
| |
| ' Nec', ' nec', |
| |
| ' nem', |
| |
| ' Nath', |
| |
| ' Hyde', |
| |
| ' Ness', |
| |
| 'orph', |
| |
| ' Param', 'param', 'Param', |
| |
| ' morph', 'morph', |
| |
| ' Kathryn', |
| |
| 'omorph', |
| |
| 'separ', |
| |
| ' Pathfinder', |
| |
| ' pathogens', |
| |
| 'opath', |
| |
| ' nurt', |
| |
| ' Bridgewater', |
| |
| ' pathology', |
| |
| 'Sym', 'sym', |
| |
| ' germ', |
| |
| ' conduit', |
| |
| ' Cavern', |
| |
| ' Pilgrim', |
| |
| ' pathological', |
| |
| 'usr', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'path-family', 1 |
|
|
| def prog_L07F08042(input_tokens): |
| """14 token families, 17 surface forms (train F1 of the uncapped optimum: 0.403). |
| Fires at 9,248 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating consequences or results""" |
| |
| fire = { |
| |
| ' result', 'result', |
| |
| ' aftermath', |
| |
| ' consequence', |
| |
| 'product', 'products', |
| |
| ' culmination', |
| |
| ' remainder', |
| |
| ' stemming', |
| |
| ' reflection', |
| |
| ' continuation', |
| |
| ' recipient', |
| |
| ' evidenced', |
| |
| ' descendants', ' descendant', |
| |
| ' grandson', |
| |
| ' brunt', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'result-family', 1 |
|
|
| def prog_L07F08043(input_tokens): |
| """35 token families, 47 surface forms (train F1 of the uncapped optimum: 0.683). |
| Fires at 11,755 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms associated with essential nutrients and their importance""" |
| |
| fire = { |
| |
| ' critical', 'critical', |
| |
| ' essential', 'essential', ' Essential', |
| |
| ' crucial', |
| |
| ' vital', ' Vital', ' vit', |
| |
| ' fundamental', |
| |
| ' essence', 'Ess', ' Essence', ' ess', |
| |
| 'ital', 'ITAL', |
| |
| ' integral', |
| |
| ' necessity', ' necessities', |
| |
| ' staple', |
| |
| ' imperative', |
| |
| ' Dion', |
| |
| 'entially', 'entials', |
| |
| 'cially', 'cial', |
| |
| ' frontline', |
| |
| ' cornerstone', |
| |
| ' invaluable', |
| |
| ' instrumental', |
| |
| ' indispensable', |
| |
| ' Anat', |
| |
| ' necess', |
| |
| ' foundational', |
| |
| ' essentials', |
| |
| 'important', |
| |
| ' paramount', |
| |
| ' extraord', |
| |
| ' revital', |
| |
| ' complementary', |
| |
| ' endemic', |
| |
| 'idential', |
| |
| ' insepar', |
| |
| ' vitality', |
| |
| ' elemental', |
| |
| ' symb', |
| |
| ' nour', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'crit-family', 1 |
|
|
| def prog_L07F08044(input_tokens): |
| """18 token families, 25 surface forms (train F1 of the uncapped optimum: 0.591). |
| Fires at 8,118 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "Far" in various contexts""" |
| |
| fire = { |
| |
| ' far', ' Far', 'Far', 'far', ' FAR', |
| |
| ' Alt', 'Alt', |
| |
| ' Farage', |
| |
| ' Cost', |
| |
| 'Dest', ' Dest', |
| |
| ' Fi', |
| |
| ' Fal', 'Fal', |
| |
| ' Farmer', |
| |
| 'FOR', |
| |
| 'hest', |
| |
| ' Dak', |
| |
| ' Cay', |
| |
| ' afar', |
| |
| ' Kov', |
| |
| ' furthe', |
| |
| ' Fou', |
| |
| 'Sab', |
| |
| ' Finder', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'far-family', 1 |
|
|
| def prog_L07F08045(input_tokens): |
| """18 token families, 23 surface forms (train F1 of the uncapped optimum: 0.817). |
| Fires at 7,779 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the name "Bill." """ |
| |
| fire = { |
| |
| ' Bill', 'Bill', ' Bills', 'bill', 'BILL', |
| |
| ' Barb', |
| |
| ' HB', 'HB', |
| |
| ' Osama', |
| |
| ' Norm', |
| |
| ' Hil', |
| |
| ' Branch', |
| |
| ' Herb', |
| |
| ' Basil', |
| |
| ' Gus', |
| |
| ' Betsy', |
| |
| 'Program', |
| |
| ' comprehens', |
| |
| 'Document', |
| |
| ' BS', |
| |
| ' Babe', |
| |
| ' Abstract', |
| |
| 'Doc', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'bill-family', 1 |
|
|
| def prog_L07F08046(input_tokens): |
| """8 token families, 8 surface forms (train F1 of the uncapped optimum: 0.082). |
| Fires at 20,879 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to performances or shows""" |
| |
| fire = { |
| |
| ' off', |
| |
| ' together', |
| |
| ' forward', |
| |
| ' aside', |
| |
| ' forth', |
| |
| ' foot', |
| |
| ' exact', |
| |
| 'ning', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'off-family', 1 |
|
|
| def prog_L07F08047(input_tokens): |
| """19 token families, 31 surface forms (train F1 of the uncapped optimum: 0.569). |
| Fires at 8,418 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to advantages or gains""" |
| |
| fire = { |
| |
| ' benefit', ' benefits', ' benefited', ' benefiting', ' Benefit', ' Benefits', 'benefit', |
| |
| ' beneficial', |
| |
| ' hurting', |
| |
| ' harmed', ' harming', ' harms', |
| |
| ' benef', 'Benef', ' Benef', 'benef', |
| |
| 'Quote', |
| |
| ' enrich', ' enriched', |
| |
| ' beneficiaries', |
| |
| ' differs', |
| |
| ' boon', |
| |
| ' disadvantaged', |
| |
| ' reap', |
| |
| ' beneficiary', |
| |
| ' aided', |
| |
| ' detriment', |
| |
| ' victimized', |
| |
| ' prosper', |
| |
| ' adversely', |
| |
| ' needy', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'benefit-family', 1 |
|
|
| def prog_L07F08048(input_tokens): |
| """23 token families, 40 surface forms (train F1 of the uncapped optimum: 0.411). |
| Fires at 28,084 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to specific days of the week, particularly weekends and Sundays""" |
| |
| fire = { |
| |
| ' days', 'days', ' Days', 'DAY', 'Days', |
| |
| ' hours', 'Hours', ' Hours', 'hours', |
| |
| ' weeks', ' Weeks', |
| |
| ' seasons', ' Seasons', |
| |
| ' nights', ' Nights', ' nightly', |
| |
| ' mornings', |
| |
| ' holidays', |
| |
| ' periods', |
| |
| ' meals', |
| |
| ' weekends', |
| |
| ' Summers', ' summers', |
| |
| ' semester', |
| |
| ' Sundays', |
| |
| ' daylight', ' Daylight', |
| |
| ' headaches', |
| |
| ' Fri', ' fri', 'Fri', |
| |
| ' Ramadan', |
| |
| ' shifts', |
| |
| ' daytime', |
| |
| ' schedules', |
| |
| ' showers', |
| |
| ' Tues', ' Tue', 'Tue', |
| |
| ' moms', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'day-family', 1 |
|
|
| def prog_L07F08049(input_tokens): |
| """22 token families, 31 surface forms (train F1 of the uncapped optimum: 0.209). |
| Fires at 261,129 of 36,786,566 positions. |
| Fires at 0.71% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: events or situations that have unexpected twists or humorous outcomes""" |
| |
| fire = { |
| |
| ' new', ' news', ' newest', ' newly', |
| |
| ' one', |
| |
| ' another', |
| |
| ' something', |
| |
| ' question', |
| |
| ' different', ' differently', |
| |
| ' thing', |
| |
| ' announce', ' announcement', ' announcements', |
| |
| ' simple', |
| |
| ' little', |
| |
| ' problem', |
| |
| ' change', |
| |
| ' changes', ' changed', |
| |
| ' interesting', ' interestingly', |
| |
| ' again', |
| |
| ' project', |
| |
| ' happened', ' happening', |
| |
| ' special', |
| |
| ' important', |
| |
| ' reason', |
| |
| ' step', |
| |
| ' message', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'new-family', 1 |
|
|
| def prog_L07F08050(input_tokens): |
| """2 token families, 2 surface forms (train F1 of the uncapped optimum: 0.308). |
| Fires at 258,591 of 36,786,566 positions. |
| Fires at 0.70% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: instances of punctuation marks, particularly commas, within a text""" |
| |
| fire = { |
| |
| ',', |
| |
| '),', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L07F08051(input_tokens): |
| """14 token families, 17 surface forms (train F1 of the uncapped optimum: 0.099). |
| Fires at 56,159 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to various categories of entities and their characteristics""" |
| |
| fire = { |
| |
| ' other', ' others', |
| |
| ' countries', |
| |
| ' groups', |
| |
| ' related', 'related', |
| |
| ' else', |
| |
| ' issues', |
| |
| ' services', |
| |
| ' devices', |
| |
| ' areas', |
| |
| ' products', 'products', |
| |
| ' items', |
| |
| ' sites', |
| |
| ' organizations', |
| |
| ' agencies', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'oth-family', 1 |
|
|
| def prog_L07F08052(input_tokens): |
| """13 token families, 18 surface forms (train F1 of the uncapped optimum: 0.133). |
| Fires at 184,095 of 36,786,566 positions. |
| Fires at 0.50% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases related to identifying individuals or communities at risk or involved in certain behaviors or situations""" |
| |
| fire = { |
| |
| ' who', |
| |
| ' people', |
| |
| ' those', |
| |
| ' man', |
| |
| ' any', |
| |
| ' suspected', ' suspects', ' suspect', |
| |
| ' children', |
| |
| ' users', ' user', |
| |
| ' person', ' persons', |
| |
| ' child', |
| |
| ' potential', |
| |
| ' terrorists', ' terrorist', |
| |
| ' woman', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'who-family', 1 |
|
|
| def prog_L07F08053(input_tokens): |
| """9 token families, 22 surface forms (train F1 of the uncapped optimum: 0.124). |
| Fires at 272,407 of 36,786,566 positions. |
| Fires at 0.74% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases that identify distinctive qualities or features""" |
| |
| fire = { |
| |
| ' that', |
| |
| ':', |
| |
| ' because', 'because', |
| |
| ' best', ' Best', 'Best', 'best', ' BEST', |
| |
| ' comes', 'Coming', 'comes', ' Comment', 'comings', |
| |
| ' 5', |
| |
| ' makes', 'makes', |
| |
| ' good', 'Good', ' GOOD', |
| |
| ' start', ' starts', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'that-family', 1 |
|
|
| def prog_L07F08054(input_tokens): |
| """28 token families, 42 surface forms (train F1 of the uncapped optimum: 0.209). |
| Fires at 95,473 of 36,786,566 positions.""" |
| |
| """prev SAE description: verbs related to commands or instructions""" |
| |
| fire = { |
| |
| ' take', 'take', |
| |
| ' go', |
| |
| ' stop', |
| |
| ' come', 'come', |
| |
| ' keep', 'keep', |
| |
| ' pay', |
| |
| ' bring', 'bring', ' Bring', |
| |
| ' move', 'move', |
| |
| ' leave', 'leave', |
| |
| ' watch', |
| |
| ' stay', 'stay', |
| |
| ' continue', 'continue', |
| |
| ' turn', |
| |
| ' remove', 'remove', |
| |
| ' consider', |
| |
| ' join', |
| |
| ' send', 'send', |
| |
| ' ask', |
| |
| ' develop', |
| |
| ' hold', |
| |
| ' learn', |
| |
| ' drop', |
| |
| ' act', |
| |
| ' investigate', |
| |
| ' pull', 'pull', |
| |
| ' improve', 'improve', |
| |
| ' seek', |
| |
| ' shut', 'shut', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'take-family', 1 |
|
|
| def prog_L07F08055(input_tokens): |
| """30 token families, 41 surface forms (train F1 of the uncapped optimum: 0.302). |
| Fires at 69,023 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to specific years and their associated statistics""" |
| |
| fire = { |
| |
| ' year', |
| |
| ' 2016', |
| |
| ' 2015', |
| |
| ' 2014', |
| |
| ' 2012', '2012', |
| |
| ' 2013', |
| |
| ' 2010', '2010', |
| |
| ' 2011', '2011', |
| |
| ' 2008', |
| |
| ' November', |
| |
| ' 2009', |
| |
| ' December', |
| |
| ' January', |
| |
| ' 2018', |
| |
| ' 2007', '2007', |
| |
| ' 2006', '2006', |
| |
| ' quarter', |
| |
| ' 2020', '2020', |
| |
| ' 2000', |
| |
| ' 2005', '2005', |
| |
| ' 1990', |
| |
| ' 2001', |
| |
| ' 1980', '1980', |
| |
| ' 2004', |
| |
| ' 2002', '2002', |
| |
| ' 2003', |
| |
| ' 2019', '2019', |
| |
| ' 1998', |
| |
| ' 1970', '1970', |
| |
| ' 1999', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'year-family', 1 |
|
|
| def prog_L07F08056(input_tokens): |
| """38 token families, 45 surface forms (train F1 of the uncapped optimum: 0.673). |
| Fires at 9,478 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to homelessness""" |
| |
| fire = { |
| |
| ' homeless', ' Homeless', |
| |
| ' shelter', ' shelters', ' Shelter', |
| |
| ' Downtown', |
| |
| 'eless', |
| |
| ' sleeping', |
| |
| ' unemployed', |
| |
| 'Hom', 'hom', |
| |
| ' gent', ' Gent', |
| |
| ' foreclosure', |
| |
| ' wilderness', ' Wilderness', |
| |
| ' euth', |
| |
| ' homelessness', |
| |
| ' Craigslist', |
| |
| ' begging', ' begg', |
| |
| ' encamp', |
| |
| ' tents', |
| |
| ' rash', |
| |
| ' Shel', |
| |
| ' disorderly', |
| |
| ' Salvation', |
| |
| 'onew', |
| |
| ' redevelopment', |
| |
| ' eviction', |
| |
| ' waterproof', |
| |
| 'obic', |
| |
| ' Samar', |
| |
| ' feral', |
| |
| ' needy', |
| |
| ' Daredevil', |
| |
| ' chronically', |
| |
| ' stigmat', |
| |
| ' Lonely', |
| |
| ' ghetto', |
| |
| ' dere', |
| |
| ' dehuman', |
| |
| 'owntown', |
| |
| 'sych', |
| |
| 'bnb', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'homeles-family', 1 |
|
|
| def prog_L07F08057(input_tokens): |
| """46 token families, 52 surface forms (train F1 of the uncapped optimum: 0.635). |
| Fires at 10,358 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to arbitration and mediation processes""" |
| |
| fire = { |
| |
| ' tribunal', ' Tribunal', |
| |
| ' medi', 'medi', |
| |
| 'oti', |
| |
| ' esc', 'escal', |
| |
| ' Neg', |
| |
| ' arbitration', |
| |
| ' foreclosure', |
| |
| ' arbit', ' Arbit', |
| |
| ' Zar', 'zar', |
| |
| ' barg', |
| |
| ' offic', |
| |
| ' trib', ' Trib', |
| |
| ' Rays', |
| |
| ' Torres', |
| |
| ' inquest', |
| |
| ' agon', |
| |
| ' liquidity', |
| |
| ' surrog', |
| |
| ' tug', |
| |
| ' Abram', |
| |
| ' radi', |
| |
| ' oscill', |
| |
| ' Jos', |
| |
| ' Naomi', |
| |
| ' coax', |
| |
| 'uez', |
| |
| ' rabb', |
| |
| ' salv', |
| |
| ' orchestr', |
| |
| ' impartial', |
| |
| ' HDMI', |
| |
| ' Syri', |
| |
| ' determin', |
| |
| 'adr', |
| |
| ' improv', |
| |
| ' Disp', |
| |
| 'atari', |
| |
| ' interf', |
| |
| ' Ire', |
| |
| ' causal', |
| |
| ' Aram', |
| |
| 'ereo', |
| |
| ' Ramsay', |
| |
| ' adjud', |
| |
| ' vic', |
| |
| ' Mou', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'tribun-family', 1 |
|
|
| def prog_L07F08058(input_tokens): |
| """20 token families, 24 surface forms (train F1 of the uncapped optimum: 0.704). |
| Fires at 6,265 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the concept of "core" as it relates to essential beliefs, elements, or values""" |
| |
| fire = { |
| |
| ' core', 'core', ' Core', 'Core', |
| |
| ' founding', 'found', |
| |
| ' flagship', |
| |
| ' hardcore', |
| |
| ' spine', |
| |
| ' essence', |
| |
| ' defining', |
| |
| ' fringe', |
| |
| ' cores', |
| |
| ' backbone', |
| |
| ' cornerstone', |
| |
| ' perimeter', |
| |
| ' baseline', |
| |
| ' hallmark', |
| |
| ' nucleus', |
| |
| ' periphery', |
| |
| ' bedrock', |
| |
| ' pillars', |
| |
| ' OG', |
| |
| ' ethos', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'core-family', 1 |
|
|
| def prog_L07F08059(input_tokens): |
| """48 token families, 53 surface forms (train F1 of the uncapped optimum: 0.710). |
| Fires at 21,508 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to a specific individual, likely someone in a notable context such as news or media""" |
| |
| fire = { |
| |
| ' golden', |
| |
| ' Charlottesville', |
| |
| 'ahl', |
| |
| ' bitcoins', ' Bitcoins', |
| |
| 'auer', |
| |
| ' Ducks', |
| |
| ' Heather', |
| |
| ' Weiner', |
| |
| ' armored', ' Armored', |
| |
| ' Mess', |
| |
| ' Hein', |
| |
| ' Treasurer', ' treasurer', |
| |
| 'ital', |
| |
| ' Erik', |
| |
| ' engulfed', ' engulf', |
| |
| ' Yosh', |
| |
| ' Alonso', |
| |
| ' Sig', |
| |
| ' hem', 'hemer', |
| |
| ' Fried', |
| |
| ' Bundy', |
| |
| 'affe', |
| |
| 'raf', |
| |
| ' Rein', |
| |
| ' eng', |
| |
| ' envelope', |
| |
| ' Meh', |
| |
| ' ledger', |
| |
| ' Das', |
| |
| ' Ashes', |
| |
| ' Abedin', |
| |
| ' Booth', |
| |
| 'orf', |
| |
| ' envelop', |
| |
| ' Horizons', |
| |
| 'affle', |
| |
| ' Axel', |
| |
| ' cerebral', |
| |
| ' Epstein', |
| |
| ' Sof', |
| |
| 'ipel', |
| |
| ' EFF', |
| |
| ' holog', |
| |
| ' Huang', |
| |
| 'eph', |
| |
| ' Neh', |
| |
| ' Dahl', |
| |
| ' Huma', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'golden-family', 1 |
|
|
| def prog_L07F08060(input_tokens): |
| """24 token families, 34 surface forms (train F1 of the uncapped optimum: 0.232). |
| Fires at 123,833 of 36,786,566 positions.""" |
| |
| """prev SAE description: concepts related to plans and objectives""" |
| |
| fire = { |
| |
| ' money', |
| |
| ' information', |
| |
| ' questions', |
| |
| ' games', |
| |
| ' data', ' DATA', |
| |
| ' things', |
| |
| ' details', |
| |
| ' jobs', |
| |
| ' photos', |
| |
| ' taxes', ' Taxes', |
| |
| ' images', |
| |
| ' ideas', |
| |
| ' products', 'Products', |
| |
| ' documents', ' Documents', |
| |
| ' pictures', |
| |
| ' millions', |
| |
| ' funds', |
| |
| ' updates', ' Updates', |
| |
| ' costs', ' Costs', |
| |
| ' drugs', |
| |
| ' resources', 'resources', |
| |
| ' items', |
| |
| ' objects', ' objectives', ' Objects', |
| |
| ' goals', ' Goals', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'money-family', 1 |
|
|
| def prog_L07F08061(input_tokens): |
| """6 token families, 10 surface forms (train F1 of the uncapped optimum: 0.097). |
| Fires at 23,048 of 36,786,566 positions.""" |
| |
| """prev SAE description: hypothetical or conditional statements""" |
| |
| fire = { |
| |
| ' were', ' Were', 'Were', 'were', |
| |
| ' wanted', |
| |
| ' weren', |
| |
| 'Had', ' Had', |
| |
| "'d", |
| |
| ' hadn', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'were-family', 1 |
|
|
| def prog_L07F08062(input_tokens): |
| """48 token families, 71 surface forms (train F1 of the uncapped optimum: 0.719). |
| Fires at 21,068 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the company Citigroup or its related entities""" |
| |
| fire = { |
| |
| ' Comcast', |
| |
| ' Equ', 'Equ', 'equ', ' EQU', |
| |
| 'por', ' Por', 'POR', |
| |
| 'Wik', ' Wik', 'wik', ' wik', |
| |
| ' Cit', ' cit', 'cit', |
| |
| ' Roberto', |
| |
| ' Luk', |
| |
| ' Schneider', |
| |
| ' Bhar', |
| |
| ' Muk', |
| |
| 'Jes', ' Jes', |
| |
| ' GE', |
| |
| ' coll', |
| |
| ' Scot', 'Scot', |
| |
| ' Volkswagen', |
| |
| ' Inst', 'INST', |
| |
| 'Cla', ' Cla', |
| |
| ' Exper', 'Exper', 'exper', |
| |
| ' dil', ' Dil', |
| |
| ' Ches', |
| |
| ' vag', ' Vag', 'vag', |
| |
| ' Barclays', |
| |
| ' Mog', ' mog', |
| |
| 'č', |
| |
| ' Levin', |
| |
| ' Vin', |
| |
| ' Cran', ' cran', |
| |
| ' Ek', |
| |
| ' Das', |
| |
| ' Aber', ' aber', |
| |
| ' Kro', |
| |
| ' Gian', |
| |
| ' Chrysler', |
| |
| ' Gig', |
| |
| ' Rubin', |
| |
| ' voc', ' Voc', |
| |
| ' Gott', |
| |
| ' Vander', |
| |
| ' revolving', |
| |
| ' encaps', |
| |
| ' Pog', |
| |
| ' Zimmer', |
| |
| ' imb', |
| |
| 'mong', |
| |
| ' Vik', |
| |
| ' upfront', |
| |
| ' plag', |
| |
| 'alg', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'comcast-family', 1 |
|
|
| def prog_L07F08063(input_tokens): |
| """16 token families, 29 surface forms (train F1 of the uncapped optimum: 0.480). |
| Fires at 12,782 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to articles or segments labeled "Also" indicating additional content or information""" |
| |
| fire = { |
| |
| ' also', ' Also', 'Also', 'also', ' ALSO', |
| |
| ' Additionally', 'Additionally', |
| |
| ' Specifically', 'Specifically', |
| |
| ' Actually', 'Actually', |
| |
| ' Especially', 'Especially', |
| |
| 'Meanwhile', |
| |
| ' Furthermore', 'Furthermore', |
| |
| ' Basically', 'Basically', |
| |
| 'Probably', ' Probably', |
| |
| 'Already', |
| |
| ' Similarly', 'Similarly', |
| |
| ' Mostly', |
| |
| ' Otherwise', 'Otherwise', |
| |
| 'Later', |
| |
| 'Elsewhere', |
| |
| ' Further', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'also-family', 1 |
|
|
| def prog_L07F08064(input_tokens): |
| """23 token families, 27 surface forms (train F1 of the uncapped optimum: 0.692). |
| Fires at 5,504 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to different models of iPads""" |
| |
| fire = { |
| |
| ' iPad', ' iPads', |
| |
| ' tablet', ' Tablet', |
| |
| ' Vita', |
| |
| ' Surface', |
| |
| ' Kindle', |
| |
| ' MacBook', |
| |
| ' iPod', |
| |
| 'Pad', |
| |
| ' Arduino', |
| |
| ' DS', |
| |
| ' Pocket', 'Pocket', |
| |
| ' PKK', |
| |
| 'ecast', |
| |
| ' gadgets', |
| |
| ' ego', |
| |
| ' grandchildren', |
| |
| ' eBook', 'ebook', |
| |
| ' ATP', |
| |
| ' Celsius', |
| |
| ' ip', |
| |
| ' DVDs', |
| |
| ' Canaveral', |
| |
| ' Dice', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ipad-family', 1 |
|
|
| def prog_L07F08065(input_tokens): |
| """17 token families, 21 surface forms (train F1 of the uncapped optimum: 0.142). |
| Fires at 12,918 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating something that stands out or is distinctive""" |
| |
| fire = { |
| |
| ' out', ' Outs', |
| |
| ' featured', ' Featured', |
| |
| ' apart', |
| |
| ' breakout', |
| |
| ' emerged', ' emerges', |
| |
| ' standout', |
| |
| ' highlighted', |
| |
| ' notable', |
| |
| 'alone', |
| |
| ' emerge', |
| |
| ' impression', |
| |
| ' differs', ' differ', |
| |
| ' premium', |
| |
| 'yp', |
| |
| ' popped', |
| |
| 'have', |
| |
| 'standing', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'out-family', 1 |
|
|
| def prog_L07F08066(input_tokens): |
| """27 token families, 39 surface forms (train F1 of the uncapped optimum: 0.622). |
| Fires at 8,936 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to transportation""" |
| |
| fire = { |
| |
| ' transport', ' Transport', ' transported', ' transporting', ' transports', ' transporter', |
| |
| ' transportation', |
| |
| 'port', 'porter', ' Ports', |
| |
| ' Dispatch', ' dispatch', |
| |
| ' Travel', 'travel', |
| |
| ' Aviation', |
| |
| ' Transit', |
| |
| ' convey', ' conveyed', |
| |
| 'loop', |
| |
| ' Toledo', |
| |
| ' whisk', |
| |
| ' Orbit', |
| |
| ' Marion', |
| |
| ' warp', ' Warp', |
| |
| ' ferry', |
| |
| ' shuttle', |
| |
| ' Weld', |
| |
| ' Transfer', 'transfer', |
| |
| ' Trave', |
| |
| ' relativity', |
| |
| ' abstraction', |
| |
| ' repatri', |
| |
| ' coff', |
| |
| ' propel', |
| |
| 'portation', |
| |
| ' relat', |
| |
| ' messenger', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'transport-family', 1 |
|
|
| def prog_L07F08067(input_tokens): |
| """2 token families, 2 surface forms (train F1 of the uncapped optimum: 0.153). |
| Fires at 15,565 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to categories or examples within a discussion""" |
| |
| fire = { |
| |
| ' like', |
| |
| ' including', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'like-family', 1 |
|
|
| def prog_L07F08068(input_tokens): |
| """35 token families, 38 surface forms (train F1 of the uncapped optimum: 0.498). |
| Fires at 14,032 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of eagerness or strong desire""" |
| |
| fire = { |
| |
| ' excited', |
| |
| ' happy', ' Happy', |
| |
| ' desperate', |
| |
| ' eager', |
| |
| ' desire', |
| |
| ' keen', |
| |
| ' determined', |
| |
| ' anxious', |
| |
| ' reluctant', |
| |
| ' vowed', |
| |
| ' curious', |
| |
| ' hungry', |
| |
| ' avid', |
| |
| ' enthusiastic', |
| |
| ' eagerly', |
| |
| ' strive', ' strives', |
| |
| ' glad', ' gladly', |
| |
| ' compelled', |
| |
| ' unwilling', |
| |
| ' scrambling', |
| |
| ' hesitant', |
| |
| ' zeal', |
| |
| ' hurry', |
| |
| ' striving', |
| |
| ' impatient', |
| |
| ' ecstatic', |
| |
| ' ardent', |
| |
| ' vying', |
| |
| ' bustling', |
| |
| ' oblivious', |
| |
| ' hast', |
| |
| 'perate', |
| |
| ' enthusiastically', |
| |
| 'interested', |
| |
| ' itching', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'excit-family', 1 |
|
|
| def prog_L07F08069(input_tokens): |
| """29 token families, 34 surface forms (train F1 of the uncapped optimum: 0.719). |
| Fires at 4,241 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of honors or awards""" |
| |
| fire = { |
| |
| ' hon', ' Hon', 'hon', 'Hon', |
| |
| ' Honor', ' honorable', |
| |
| ' Lauren', |
| |
| ' Fav', |
| |
| ' FY', 'FY', |
| |
| ' honorary', |
| |
| 'pri', |
| |
| ' Duc', |
| |
| ' Friend', |
| |
| ' dishon', |
| |
| 'WIN', |
| |
| 'san', |
| |
| 'ardon', |
| |
| ' unquestion', |
| |
| ' Holden', |
| |
| ' Mour', |
| |
| ' patriotism', |
| |
| 'heed', |
| |
| ' unsur', |
| |
| ' Frazier', |
| |
| 'achel', |
| |
| ' Honour', |
| |
| ' POS', |
| |
| 'Exc', |
| |
| ' Nos', |
| |
| ' laud', |
| |
| ' Favor', |
| |
| 'Spons', |
| |
| ' regrett', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'hon-family', 1 |
|
|
| def prog_L07F08070(input_tokens): |
| """1 token families, 5 surface forms (train F1 of the uncapped optimum: 0.387). |
| Fires at 11,279 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of uncertainty or lack of knowledge in the text""" |
| |
| fire = { |
| |
| ' know', ' knows', ' knowing', 'know', ' KNOW', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'know-family', 1 |
|
|
| def prog_L07F08071(input_tokens): |
| """49 token families, 55 surface forms (train F1 of the uncapped optimum: 0.675). |
| Fires at 10,502 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "nod" and its variations indicating agreement or acknowledgment""" |
| |
| fire = { |
| |
| ' NDP', |
| |
| ' asleep', |
| |
| ' nod', ' nods', |
| |
| ' Nigel', |
| |
| ' nominations', |
| |
| ' sigh', ' sighed', |
| |
| ' bribery', |
| |
| ' nephew', |
| |
| ' bribes', |
| |
| ' thumbs', |
| |
| 'Yep', ' Yep', |
| |
| ' Sonia', |
| |
| ' Naw', |
| |
| ' succumbed', ' succumb', |
| |
| ' bob', |
| |
| ' cough', ' coughing', |
| |
| ' homage', |
| |
| ' concession', |
| |
| ' Redmond', |
| |
| ' concessions', |
| |
| ' cheers', |
| |
| ' openness', |
| |
| ' quasi', |
| |
| ' yen', |
| |
| ' Nou', |
| |
| ' noun', |
| |
| ' scoff', |
| |
| ' blink', ' blinked', |
| |
| ' nud', |
| |
| ' euphem', |
| |
| ' Pok', |
| |
| ' niece', |
| |
| ' sleepy', |
| |
| ' bowed', |
| |
| ' salute', |
| |
| ' Nope', |
| |
| 'umbs', |
| |
| ' tacit', |
| |
| ' shrug', |
| |
| ' smug', |
| |
| ' limp', |
| |
| ' paternal', |
| |
| ' Greenberg', |
| |
| ' shrugged', |
| |
| ' yawn', |
| |
| ' Nay', |
| |
| 'udging', |
| |
| ' attest', |
| |
| 'uppet', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ndp-family', 1 |
|
|
| def prog_L07F08072(input_tokens): |
| """8 token families, 13 surface forms (train F1 of the uncapped optimum: 0.821). |
| Fires at 30,836 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of law enforcement or police activity""" |
| |
| fire = { |
| |
| ' police', ' Police', 'Police', |
| |
| 'Authorities', ' Authorities', |
| |
| 'Investigators', ' Investigators', |
| |
| 'Officers', |
| |
| ' RCMP', |
| |
| ' detectives', 'Detect', |
| |
| ' TMZ', |
| |
| 'CTV', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'police-family', 1 |
|
|
| def prog_L07F08073(input_tokens): |
| """7 token families, 13 surface forms (train F1 of the uncapped optimum: 0.077). |
| Fires at 237,284 of 36,786,566 positions. |
| Fires at 0.65% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: specific events, actions, or subjects that have numeric or temporal significance""" |
| |
| fire = { |
| |
| '�', |
| |
| ' I', ' i', |
| |
| ' Image', 'image', |
| |
| ' Media', |
| |
| ' A', 'ça', |
| |
| ' President', 'President', |
| |
| ' We', 'WE', ' WE', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '�-family', 1 |
|
|
| def prog_L07F08074(input_tokens): |
| """3 token families, 3 surface forms (train F1 of the uncapped optimum: 0.150). |
| Fires at 75,008 of 36,786,566 positions.""" |
| |
| """prev SAE description: connections between policies and their impacts on various societal aspects""" |
| |
| fire = { |
| |
| ' and', |
| |
| ' or', |
| |
| ' expressed', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'and-family', 1 |
|
|
| def prog_L07F08075(input_tokens): |
| """23 token families, 36 surface forms (train F1 of the uncapped optimum: 0.879). |
| Fires at 18,288 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to specific dates and numerical data""" |
| |
| fire = { |
| |
| ' 15', '15', |
| |
| ' 150', '150', |
| |
| ' 55', |
| |
| 'teen', |
| |
| ' fifteen', |
| |
| ' 1500', '1500', |
| |
| ' 155', '155', |
| |
| ' 1600', '1600', |
| |
| ' 1915', |
| |
| ' Ches', |
| |
| ' Sag', |
| |
| ' 1910', |
| |
| ' 152', '152', |
| |
| ' Continental', |
| |
| ' 153', '153', |
| |
| ' 158', '158', |
| |
| ' Lexington', |
| |
| ' 151', '151', |
| |
| '315', ' 315', |
| |
| ' 157', '157', |
| |
| ' 156', '156', |
| |
| ' 159', '159', |
| |
| 'irteen', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '15-family', 1 |
|
|
| def prog_L07F08076(input_tokens): |
| """50 token families, 60 surface forms (train F1 of the uncapped optimum: 0.655). |
| Fires at 29,946 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of JavaScript and related web technology terms""" |
| |
| fire = { |
| |
| ' JavaScript', ' javascript', 'javascript', |
| |
| 'umble', |
| |
| 'ender', |
| |
| ' extinction', |
| |
| ' unlocks', |
| |
| ' Pastebin', |
| |
| ' Reader', 'readable', |
| |
| 'iper', 'ipers', |
| |
| ' penned', |
| |
| ' fairy', ' Fairy', |
| |
| ' Noble', |
| |
| ' Print', |
| |
| ' undergraduate', |
| |
| ' batting', ' batted', |
| |
| ' qualifier', ' qualifiers', |
| |
| ' Robbie', |
| |
| ' Abdul', |
| |
| 'chie', |
| |
| ' Safari', |
| |
| ' Motorsport', |
| |
| ' Stub', ' stub', |
| |
| 'jee', |
| |
| ' Lucy', |
| |
| ' Nim', |
| |
| ' Vanity', ' vanity', |
| |
| ' Carly', |
| |
| 'phy', |
| |
| ' Sheldon', |
| |
| ' Trophy', |
| |
| ' Fletcher', |
| |
| ' Orwell', |
| |
| ' understandable', |
| |
| 'ully', |
| |
| ' Wembley', |
| |
| ' Pearce', |
| |
| ' Albion', |
| |
| ' Etsy', 'etsy', |
| |
| ' Elliot', |
| |
| ' turf', |
| |
| ' Coulter', |
| |
| ' Sharia', |
| |
| ' Booth', |
| |
| ' McCoy', |
| |
| ' goalie', |
| |
| ' Jag', |
| |
| 'obia', |
| |
| ' MacDonald', |
| |
| ' gunned', |
| |
| ' falsehood', |
| |
| ' Alec', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'javascript-family', 1 |
|
|
| def prog_L07F08077(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.439). |
| Fires at 63,396 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the letter 's'""" |
| |
| fire = { |
| |
| 's', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 's-family', 1 |
|
|
| def prog_L07F08078(input_tokens): |
| """9 token families, 31 surface forms (train F1 of the uncapped optimum: 0.119). |
| Fires at 345,058 of 36,786,566 positions. |
| Fires at 0.94% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases related to community engagement and interaction""" |
| |
| fire = { |
| |
| ' and', 'anding', |
| |
| ' their', 'their', |
| |
| ' more', |
| |
| ' people', |
| |
| ' our', |
| |
| ' your', 'your', ' yours', ' YOUR', |
| |
| ' development', ' develop', ' developing', ' developers', ' Development', 'develop', ' develops', 'development', ' Develop', |
| |
| ' world', ' worlds', ' worldly', |
| |
| ' community', ' communities', ' commun', ' Communities', 'community', ' communal', 'Commun', ' Commun', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'and-family', 1 |
|
|
| def prog_L07F08079(input_tokens): |
| """11 token families, 24 surface forms (train F1 of the uncapped optimum: 0.090). |
| Fires at 246,871 of 36,786,566 positions. |
| Fires at 0.67% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases related to websites and online presence""" |
| |
| fire = { |
| |
| ',', ' ,', |
| |
| ' has', ' HAS', |
| |
| ' which', 'which', |
| |
| ' here', ' Here', 'here', ' HERE', |
| |
| ' where', 'where', |
| |
| 'com', |
| |
| ' so', |
| |
| ' posted', ' posts', ' posting', ' postings', |
| |
| ' site', ' Site', 'site', |
| |
| ' website', ' Website', |
| |
| ' called', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L07F08080(input_tokens): |
| """16 token families, 23 surface forms (train F1 of the uncapped optimum: 0.490). |
| Fires at 6,755 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to expressions of concern or care""" |
| |
| fire = { |
| |
| ' care', 'Care', ' CARE', |
| |
| ' cares', ' caring', ' cared', |
| |
| ' worry', ' worrying', |
| |
| ' dreamed', ' dreaming', |
| |
| ' depend', 'depend', |
| |
| 'assion', |
| |
| ' reck', |
| |
| ' cater', |
| |
| ' compassion', |
| |
| ' mattered', |
| |
| ' disappro', |
| |
| ' christ', |
| |
| 'iem', |
| |
| 'icion', |
| |
| ' indifferent', |
| |
| ' rooting', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'care-family', 1 |
|
|
| def prog_L07F08081(input_tokens): |
| """44 token families, 46 surface forms (train F1 of the uncapped optimum: 0.630). |
| Fires at 9,408 of 36,786,566 positions.""" |
| |
| """prev SAE description: the term "consistent" and its derivatives in various contexts""" |
| |
| fire = { |
| |
| ' consistent', ' consistently', |
| |
| ' persistent', |
| |
| ' predictable', |
| |
| ' continuous', ' continuity', |
| |
| ' unanimous', |
| |
| ' consistency', |
| |
| ' comparable', |
| |
| ' accordance', |
| |
| 'istent', |
| |
| ' inconsistent', |
| |
| ' systematic', |
| |
| ' reliably', |
| |
| ' coherent', |
| |
| ' erratic', |
| |
| ' parity', |
| |
| ' unified', |
| |
| ' unchanged', |
| |
| ' uneven', |
| |
| ' predictably', |
| |
| ' collegiate', |
| |
| ' sloppy', |
| |
| ' consensual', |
| |
| ' discrepancies', |
| |
| 'herent', |
| |
| ' contradictory', |
| |
| ' blatant', |
| |
| ' discrepancy', |
| |
| ' inconsistencies', |
| |
| 'avering', |
| |
| ' conflicting', |
| |
| ' sporadic', |
| |
| ' incompatible', |
| |
| ' cohesive', |
| |
| ' disparity', |
| |
| ' uniformly', |
| |
| 'uity', |
| |
| ' inconsistency', |
| |
| ' deviation', |
| |
| ' concurrent', |
| |
| 'idespread', |
| |
| 'herence', |
| |
| ' uninterrupted', |
| |
| ' variance', |
| |
| 'etrical', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'consistent-family', 1 |
|
|
| def prog_L07F08082(input_tokens): |
| """15 token families, 19 surface forms (train F1 of the uncapped optimum: 0.657). |
| Fires at 5,551 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "let." """ |
| |
| fire = { |
| |
| 'let', 'lets', 'LET', ' LET', |
| |
| 'rot', |
| |
| 'hya', |
| |
| 'plet', 'plets', |
| |
| 'letico', |
| |
| 'lest', |
| |
| 'jet', |
| |
| ' booklet', |
| |
| 'pour', |
| |
| 'ourt', |
| |
| ' clot', |
| |
| ' hepatitis', |
| |
| 'format', |
| |
| 'POR', |
| |
| ' knit', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'let-family', 1 |
|
|
| def prog_L07F08083(input_tokens): |
| """45 token families, 61 surface forms (train F1 of the uncapped optimum: 0.657). |
| Fires at 11,095 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the term "Rav" and its variations, likely indicating a focus on specific individuals or entities associated with that name""" |
| |
| fire = { |
| |
| 'rav', ' Rav', ' rav', 'ravity', |
| |
| 'lev', 'Lev', |
| |
| ' aggravated', |
| |
| 'urt', |
| |
| ' Nev', 'Nev', |
| |
| 'ava', |
| |
| 'umm', |
| |
| 'riv', ' Riv', |
| |
| ' SV', 'sv', ' sv', |
| |
| 'ament', |
| |
| 'urb', |
| |
| ' grav', ' Grav', |
| |
| 'uv', 'UV', |
| |
| ' Vital', ' vit', |
| |
| ' Dust', 'Dust', |
| |
| ' Jorge', |
| |
| 'lv', 'LV', ' LV', |
| |
| 'č', |
| |
| 'rave', |
| |
| 'Av', |
| |
| 'ripp', |
| |
| 'rov', |
| |
| ' Kov', |
| |
| 'rev', |
| |
| ' IPv', |
| |
| 'atri', |
| |
| 'orge', |
| |
| ' Hurt', |
| |
| ' Taj', |
| |
| ' Bravo', |
| |
| ' rive', |
| |
| ' Thornton', |
| |
| ' Hav', |
| |
| ' Eucl', |
| |
| ' Lov', |
| |
| 'igor', |
| |
| ' Scorp', |
| |
| ' RGB', |
| |
| ' Mov', |
| |
| ' Stev', |
| |
| ' VAL', 'VAL', |
| |
| ' Pavel', |
| |
| ' buoy', |
| |
| 'olver', |
| |
| ' VK', 'VK', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'rav-family', 1 |
|
|
| def prog_L07F08084(input_tokens): |
| """19 token families, 26 surface forms (train F1 of the uncapped optimum: 0.210). |
| Fires at 28,372 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of phrases that indicate first impressions or appearances""" |
| |
| fire = { |
| |
| ' surface', |
| |
| ' eye', ' eyes', |
| |
| ' beginning', ' beginnings', |
| |
| ' itself', |
| |
| ' glance', |
| |
| 'term', |
| |
| ' core', |
| |
| ' distance', |
| |
| 'hand', 'handed', |
| |
| ' observers', ' observer', |
| |
| ' inception', |
| |
| ' trail', |
| |
| ' visually', |
| |
| ' outset', |
| |
| ' innocuous', |
| |
| ' radar', ' Radar', |
| |
| ' intuitive', 'intuitive', |
| |
| 'stock', |
| |
| ' outsiders', ' outsider', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'surface-family', 1 |
|
|
| def prog_L07F08085(input_tokens): |
| """4 token families, 5 surface forms (train F1 of the uncapped optimum: 0.061). |
| Fires at 45,439 of 36,786,566 positions.""" |
| |
| """prev SAE description: instructions related to cooking and food preparation""" |
| |
| fire = { |
| |
| ' or', |
| |
| ' keep', ' keeps', |
| |
| ' until', |
| |
| ' every', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'or-family', 1 |
|
|
| def prog_L07F08086(input_tokens): |
| """13 token families, 28 surface forms (train F1 of the uncapped optimum: 0.070). |
| Fires at 220,121 of 36,786,566 positions. |
| Fires at 0.60% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: specific names and terms related to political and societal issues""" |
| |
| fire = { |
| |
| ' a', |
| |
| ' �', |
| |
| ' all', 'All', 'all', ' ALL', ' allied', |
| |
| ' one', ' ONE', 'onement', |
| |
| ' there', 'there', |
| |
| ' Trump', ' trump', |
| |
| ' people', 'people', ' PEOPLE', |
| |
| " '", "'", |
| |
| ' two', ' TWO', |
| |
| ' some', 'some', |
| |
| ' many', 'many', |
| |
| ' those', 'those', |
| |
| ' such', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'a-family', 1 |
|
|
| def prog_L07F08087(input_tokens): |
| """40 token families, 61 surface forms (train F1 of the uncapped optimum: 0.426). |
| Fires at 24,470 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to Kickstarter campaigns and related terminology""" |
| |
| fire = { |
| |
| ' campaign', 'Campaign', |
| |
| ' funding', ' funded', 'funded', 'Fund', 'funding', 'fund', ' Funding', |
| |
| ' Kickstarter', |
| |
| 'order', 'orders', 'Order', |
| |
| ' backers', ' backer', |
| |
| ' Steam', |
| |
| ' fundraising', ' fundraiser', |
| |
| ' ICO', 'ICO', |
| |
| ' crowdfunding', |
| |
| ' beta', ' Beta', |
| |
| ' grassroots', |
| |
| ' stretch', ' Stretch', 'Stretch', |
| |
| ' pledge', ' Pledge', |
| |
| ' petition', |
| |
| ' Kick', 'Kick', |
| |
| 'starter', ' Starter', 'started', |
| |
| ' PayPal', |
| |
| ' DRM', |
| |
| ' Early', |
| |
| ' donation', |
| |
| 'ourced', 'ourcing', |
| |
| ' subscription', |
| |
| ' countdown', |
| |
| ' ransom', |
| |
| ' PAX', |
| |
| ' collectors', ' Collector', |
| |
| ' Citizen', |
| |
| 'ogo', |
| |
| ' Patreon', |
| |
| ' tabletop', |
| |
| 'ETA', |
| |
| ' Indie', |
| |
| ' Arduino', |
| |
| ' Pebble', |
| |
| 'kamp', |
| |
| ' pledges', |
| |
| 'utra', |
| |
| 'ledge', |
| |
| ' CGI', |
| |
| 'release', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'campaign-family', 1 |
|
|
| def prog_L07F08088(input_tokens): |
| """8 token families, 13 surface forms (train F1 of the uncapped optimum: 0.772). |
| Fires at 10,223 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to various types of products""" |
| |
| fire = { |
| |
| ' products', ' product', ' Products', 'products', ' PRODUCT', 'Products', |
| |
| ' merchandise', |
| |
| ' creations', |
| |
| ' marketed', |
| |
| ' supplements', |
| |
| ' cosmetics', |
| |
| ' merch', |
| |
| ' inventions', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'product-family', 1 |
|
|
| def prog_L07F08089(input_tokens): |
| """3 token families, 10 surface forms (train F1 of the uncapped optimum: 0.389). |
| Fires at 79,792 of 36,786,566 positions.""" |
| |
| """prev SAE description: possessive pronouns and their associations""" |
| |
| fire = { |
| |
| ' its', ' Its', 'Its', ' ITS', |
| |
| ' their', ' Their', 'Their', 'their', |
| |
| ' whose', 'whose', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'its-family', 1 |
|
|
| def prog_L07F08090(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.239). |
| Fires at 31,073 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "it" in various contexts, indicating a focus on pronouns""" |
| |
| fire = { |
| |
| ' it', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'it-family', 1 |
|
|
| def prog_L07F08091(input_tokens): |
| """27 token families, 38 surface forms (train F1 of the uncapped optimum: 0.616). |
| Fires at 10,284 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to avoidance""" |
| |
| fire = { |
| |
| ' avoid', ' avoided', ' avoiding', ' avoidance', ' Avoid', ' avoids', 'Avoid', 'avoid', |
| |
| ' prevent', |
| |
| ' dodged', ' dodging', |
| |
| ' minimize', |
| |
| ' aver', ' aversion', |
| |
| ' wary', |
| |
| ' spared', |
| |
| ' Dai', |
| |
| ' mitigate', |
| |
| ' beware', ' Beware', |
| |
| ' shun', |
| |
| ' dodge', |
| |
| ' avert', |
| |
| ' evade', |
| |
| ' esche', |
| |
| ' preempt', |
| |
| ' refrain', |
| |
| ' circumvent', |
| |
| ' omitted', |
| |
| ' unavoidable', ' unavoid', |
| |
| ' anticipate', |
| |
| ' fearing', |
| |
| ' eliminates', |
| |
| ' deflect', |
| |
| ' minimizing', |
| |
| ' anticipating', |
| |
| 'eware', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'avoid-family', 1 |
|
|
| def prog_L07F08092(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.294). |
| Fires at 97,064 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "in" used as a preposition""" |
| |
| fire = { |
| |
| ' in', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'in-family', 1 |
|
|
| def prog_L07F08093(input_tokens): |
| """37 token families, 48 surface forms (train F1 of the uncapped optimum: 0.595). |
| Fires at 12,793 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of "waters," implying a focus on marine or aquatic contexts""" |
| |
| fire = { |
| |
| ' water', ' waters', 'water', ' Waters', 'waters', |
| |
| ' waves', ' Waves', |
| |
| ' seas', ' Seas', |
| |
| ' underwater', |
| |
| ' skies', |
| |
| ' rivers', ' Rivers', |
| |
| ' rocks', ' Rocks', |
| |
| ' shores', |
| |
| 'aters', 'ATER', |
| |
| ' whales', |
| |
| ' boats', 'boats', |
| |
| ' depths', |
| |
| ' oceans', |
| |
| ' washed', |
| |
| ' airspace', |
| |
| ' vessels', |
| |
| ' disputed', |
| |
| ' maritime', |
| |
| ' murky', |
| |
| ' watershed', |
| |
| ' waterways', |
| |
| ' Strait', |
| |
| ' sailed', ' sails', |
| |
| ' submerged', |
| |
| ' rocky', |
| |
| ' afloat', |
| |
| ' waterfront', |
| |
| ' fray', |
| |
| ' warships', |
| |
| ' salty', |
| |
| ' reefs', |
| |
| ' Burgess', |
| |
| ' Kuh', |
| |
| ' wetlands', |
| |
| ' fluids', |
| |
| ' overboard', |
| |
| 'ipples', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'wat-family', 1 |
|
|
| def prog_L07F08094(input_tokens): |
| """6 token families, 16 surface forms (train F1 of the uncapped optimum: 0.865). |
| Fires at 12,994 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to "data" in various contexts related to technology and information processing""" |
| |
| fire = { |
| |
| ' data', ' Data', 'Data', 'data', ' DATA', 'DATA', ' datas', |
| |
| ' Dat', ' dat', 'Dat', 'dat', |
| |
| 'abytes', |
| |
| ' dataset', ' datasets', |
| |
| 'ー', |
| |
| 'metry', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'data-family', 1 |
|
|
| def prog_L07F08095(input_tokens): |
| """2 token families, 2 surface forms (train F1 of the uncapped optimum: 0.172). |
| Fires at 115,817 of 36,786,566 positions.""" |
| |
| """prev SAE description: sentences or phrases that provide conclusions or summaries""" |
| |
| fire = { |
| |
| '.', |
| |
| ').', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '.-family', 1 |
|
|
| def prog_L07F08096(input_tokens): |
| """6 token families, 7 surface forms (train F1 of the uncapped optimum: 0.705). |
| Fires at 10,094 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the letter 'b'""" |
| |
| fire = { |
| |
| 'b', ' b', |
| |
| ' bu', |
| |
| ' bro', |
| |
| ' bun', |
| |
| ' ber', |
| |
| ' biking', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'b-family', 1 |
|
|
| def prog_L07F08097(input_tokens): |
| """25 token families, 33 surface forms (train F1 of the uncapped optimum: 0.193). |
| Fires at 141,622 of 36,786,566 positions.""" |
| |
| """prev SAE description: experiences of personal struggles and achievements""" |
| |
| fire = { |
| |
| ' have', |
| |
| ' had', |
| |
| "'m", 'm', |
| |
| "'ve", 've', |
| |
| ' am', |
| |
| "'t", |
| |
| ' don', |
| |
| ' want', ' wanted', |
| |
| ' got', |
| |
| 'll', "'ll", |
| |
| ' didn', 'didn', |
| |
| ' won', |
| |
| ' did', ' DID', |
| |
| ' always', ' ALWAYS', |
| |
| ' never', |
| |
| ' decided', |
| |
| ' started', |
| |
| ' couldn', |
| |
| ' feel', |
| |
| "'d", |
| |
| ' love', ' LOVE', |
| |
| ' knew', |
| |
| ' haven', |
| |
| ' wasn', |
| |
| ' spent', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'have-family', 1 |
|
|
| def prog_L07F08098(input_tokens): |
| """34 token families, 40 surface forms (train F1 of the uncapped optimum: 0.669). |
| Fires at 6,235 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to lakes and their related features""" |
| |
| fire = { |
| |
| ' Lake', ' lake', 'lake', 'Lake', |
| |
| ' Lakes', ' lakes', |
| |
| ' Basin', ' basin', |
| |
| ' Baltic', |
| |
| ' reservoirs', ' reservoir', |
| |
| ' algae', |
| |
| 'amin', |
| |
| ' dil', |
| |
| ' Dai', |
| |
| ' Erie', |
| |
| ' overlooking', |
| |
| ' shallow', |
| |
| ' almond', |
| |
| ' Kobe', |
| |
| ' harbour', |
| |
| ' Darling', |
| |
| 'apeake', |
| |
| 'azaki', |
| |
| ' breasts', |
| |
| ' lobster', |
| |
| ' wetlands', |
| |
| ' Axel', |
| |
| ' ponds', |
| |
| ' freshwater', |
| |
| ' amino', |
| |
| 'irie', |
| |
| ' lob', |
| |
| 'hai', |
| |
| ' Shad', |
| |
| ' retri', |
| |
| ' Admir', |
| |
| ' blob', |
| |
| ' mars', |
| |
| ' amph', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'lake-family', 1 |
|
|
| def prog_L07F08099(input_tokens): |
| """32 token families, 61 surface forms (train F1 of the uncapped optimum: 0.171). |
| Fires at 200,709 of 36,786,566 positions. |
| Fires at 0.55% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: words associated with leaks, spills, or discharges in various contexts""" |
| |
| fire = { |
| |
| '-', |
| |
| 't', 'ét', |
| |
| 'th', 'TH', |
| |
| 'a', 'ía', |
| |
| 'i', |
| |
| 'o', 'ão', 'ño', |
| |
| 'as', 'AS', |
| |
| 'y', 'Y', |
| |
| 'ar', 'AR', |
| |
| 'an', 'AN', |
| |
| 'z', 'Z', |
| |
| 'on', 'ON', |
| |
| 'n', 'ön', 'én', 'án', 'ón', 'ín', |
| |
| 'en', 'EN', |
| |
| '0', |
| |
| 'al', 'AL', |
| |
| 'is', |
| |
| 'am', 'AM', |
| |
| 'at', 'AT', |
| |
| 'r', 'ür', 'ör', 'ré', |
| |
| 'u', |
| |
| 'le', 'LE', |
| |
| 'h', |
| |
| 'e', |
| |
| 'p', |
| |
| 'um', 'UM', |
| |
| 'k', |
| |
| 'v', |
| |
| 'ic', |
| |
| 'ch', 'CH', |
| |
| 'ine', 'ines', 'INE', |
| |
| 'id', 'ID', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '--family', 1 |
|
|
