| """GPT-2 SAE last-token programs, layer 6 -- features 21900 to 21999 (100 programs). |
| |
| One function per SAE feature. Grouped 100 to a file so the set stays browsable; |
| the programs themselves are unchanged. |
| """ |
|
|
| def prog_L06F21900(input_tokens): |
| """4 token families, 10 surface forms (train F1 of the uncapped optimum: 0.656). |
| Fires at 22,854 of 36,786,566 positions.""" |
| |
| """prev SAE description: instructions for utilizing various tools or techniques""" |
| |
| fire = { |
| |
| ' use', ' uses', ' Use', 'Use', ' Uses', ' USE', ' Useful', |
| |
| ' utilize', |
| |
| ' employ', |
| |
| ' usable', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'use-family', 1 |
|
|
| def prog_L06F21901(input_tokens): |
| """12 token families, 12 surface forms (train F1 of the uncapped optimum: 0.207). |
| Fires at 13,088 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to a specific entity or organization""" |
| |
| fire = { |
| |
| 'C', |
| |
| ' Columbia', |
| |
| '#', |
| |
| ' Budget', |
| |
| 'UL', |
| |
| 'ali', |
| |
| 'oca', |
| |
| 'band', |
| |
| 'payer', |
| |
| 'ibi', |
| |
| 'aca', |
| |
| 'uru', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'c-family', 1 |
|
|
| def prog_L06F21902(input_tokens): |
| """19 token families, 19 surface forms (train F1 of the uncapped optimum: 0.156). |
| Fires at 23,500 of 36,786,566 positions.""" |
| |
| """prev SAE description: variations of the prefix "und" in words""" |
| |
| fire = { |
| |
| 'en', |
| |
| 'of', |
| |
| 'ig', |
| |
| 'ul', |
| |
| 'ew', |
| |
| 'ET', |
| |
| 'é', |
| |
| 'RO', |
| |
| 'IAN', |
| |
| 'ere', |
| |
| 'ef', |
| |
| 'pre', |
| |
| ' der', |
| |
| 'des', |
| |
| 'eter', |
| |
| 'ieg', |
| |
| 'esc', |
| |
| 'erm', |
| |
| 'ener', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'en-family', 1 |
|
|
| def prog_L06F21903(input_tokens): |
| """17 token families, 17 surface forms (train F1 of the uncapped optimum: 0.106). |
| Fires at 163,372 of 36,786,566 positions. |
| Fires at 0.44% 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: occurrences of specific days and actions related to meetings or interactions""" |
| |
| fire = { |
| |
| ' on', |
| |
| ' at', |
| |
| ' after', |
| |
| ' during', |
| |
| ' before', |
| |
| ' since', |
| |
| ' while', |
| |
| ' Wednesday', |
| |
| ' Tuesday', |
| |
| ' Monday', |
| |
| ' today', |
| |
| ' Thursday', |
| |
| ' earlier', |
| |
| ' following', |
| |
| ' morning', |
| |
| ' via', |
| |
| ' discuss', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'on-family', 1 |
|
|
| def prog_L06F21904(input_tokens): |
| """38 token families, 44 surface forms (train F1 of the uncapped optimum: 0.816). |
| Fires at 6,108 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the term "Ter" in various contexts""" |
| |
| fire = { |
| |
| ' Ter', ' ter', 'Ter', ' TER', |
| |
| ' Tur', |
| |
| ' Zimbabwe', |
| |
| ' terrain', |
| |
| ' Tier', 'Tier', |
| |
| ' Trafford', |
| |
| ' terr', ' Terr', |
| |
| 'Mor', |
| |
| ' Nut', |
| |
| ' tor', |
| |
| ' Vers', |
| |
| ' uproar', |
| |
| ' extrater', |
| |
| ' Mist', |
| |
| ' Tet', |
| |
| ' Tig', |
| |
| ' sanct', |
| |
| ' Terra', |
| |
| 'Termin', ' Termin', |
| |
| ' tact', |
| |
| ' Spur', |
| |
| ' subter', |
| |
| ' Ster', |
| |
| 'ン', |
| |
| ' Zer', |
| |
| ' Vor', |
| |
| ' retard', |
| |
| ' teaspoon', |
| |
| ' Territories', |
| |
| ' Alter', |
| |
| ' Tus', |
| |
| ' ont', |
| |
| ' Terran', |
| |
| ' etiquette', |
| |
| 'Palest', |
| |
| 'imbabwe', |
| |
| ' Stellar', |
| |
| ' Territ', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ter-family', 1 |
|
|
| def prog_L06F21905(input_tokens): |
| """12 token families, 16 surface forms (train F1 of the uncapped optimum: 0.156). |
| Fires at 28,901 of 36,786,566 positions.""" |
| |
| """prev SAE description: emotional states related to safety, comfort, and feelings of guilt or exclusion""" |
| |
| fire = { |
| |
| ' comfortable', ' comfort', ' comforting', |
| |
| ' safe', 'safe', |
| |
| ' dead', |
| |
| 'bad', |
| |
| ' confident', |
| |
| ' compelled', |
| |
| ' uncomfortable', |
| |
| ' sad', ' sadness', |
| |
| ' honored', |
| |
| ' embarrassed', |
| |
| ' vulnerable', |
| |
| ' lucky', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'comfort-family', 1 |
|
|
| def prog_L06F21906(input_tokens): |
| """17 token families, 19 surface forms (train F1 of the uncapped optimum: 0.182). |
| Fires at 10,182 of 36,786,566 positions.""" |
| |
| """prev SAE description: profanity and explicit language""" |
| |
| fire = { |
| |
| '*', |
| |
| '***', |
| |
| '**', |
| |
| '#', |
| |
| 'ü', |
| |
| ' heck', |
| |
| 'odd', |
| |
| '****', ' ****', |
| |
| '---', |
| |
| 'url', |
| |
| 'other', |
| |
| '??', ' ??', |
| |
| 'ø', |
| |
| ')(', |
| |
| 'pic', |
| |
| 'ê', |
| |
| '¢', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '*-family', 1 |
|
|
| def prog_L06F21907(input_tokens): |
| """12 token families, 19 surface forms (train F1 of the uncapped optimum: 0.459). |
| Fires at 7,935 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "reporting." """ |
| |
| fire = { |
| |
| ' reporting', ' Reporting', 'reported', 'reporting', 'report', ' REPORT', 'reports', 'Reporting', |
| |
| ' storytelling', |
| |
| ' Rapp', |
| |
| ' organizing', |
| |
| ' journalistic', |
| |
| ' Battalion', |
| |
| 'checking', |
| |
| ' piping', |
| |
| 'icing', |
| |
| ' Braz', |
| |
| 'dropping', |
| |
| 'telling', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'report-family', 1 |
|
|
| def prog_L06F21908(input_tokens): |
| """44 token families, 102 surface forms (train F1 of the uncapped optimum: 0.421). |
| Fires at 85,447 of 36,786,566 positions.""" |
| |
| """prev SAE description: technical terms related to audio processing and monitoring""" |
| |
| fire = { |
| |
| 'ing', |
| |
| ' management', ' Management', 'management', 'Management', |
| |
| ' testing', ' Testing', 'testing', 'Testing', ' TEST', |
| |
| ' search', ' Search', 'search', 'Search', |
| |
| ' marketing', ' Marketing', |
| |
| ' analysis', ' Analysis', |
| |
| ' payment', ' payments', ' Payments', ' Payment', 'payment', |
| |
| ' sharing', 'sharing', ' Sharing', |
| |
| ' surveillance', ' Surveillance', |
| |
| 'ation', |
| |
| ' processing', ' Processing', 'processing', |
| |
| 'ization', |
| |
| ' tracking', 'tracking', ' Tracking', |
| |
| ' learning', ' Learning', 'learning', |
| |
| ' parking', ' Parking', |
| |
| ' storage', ' Storage', 'storage', |
| |
| ' mining', ' Mining', 'mining', |
| |
| ' programming', ' Programming', |
| |
| ' voting', ' Voting', |
| |
| ' streaming', ' Streaming', 'Streamer', |
| |
| ' trading', ' Trading', |
| |
| ' combat', ' Combat', |
| |
| ' communication', 'communication', ' Communication', |
| |
| ' advertising', ' Advertising', 'advertising', |
| |
| ' monitoring', ' Monitoring', |
| |
| ' printing', ' Printing', |
| |
| ' computing', ' Computing', |
| |
| ' delivery', ' Delivery', |
| |
| ' manufacturing', ' Manufacturing', |
| |
| ' shipping', ' Shipping', 'Shipping', |
| |
| ' lighting', ' Lighting', 'lighting', |
| |
| ' dating', ' Dating', 'dating', |
| |
| ' messaging', ' Messages', |
| |
| ' encryption', |
| |
| ' communications', 'communications', |
| |
| ' charging', 'charging', |
| |
| ' navigation', ' Navigation', |
| |
| ' graphics', |
| |
| ' imaging', ' Imaging', |
| |
| ' detection', ' Detection', |
| |
| 'omy', |
| |
| ' recognition', |
| |
| ' analytics', ' Analytics', |
| |
| 'ification', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ing-family', 1 |
|
|
| def prog_L06F21909(input_tokens): |
| """12 token families, 17 surface forms (train F1 of the uncapped optimum: 0.745). |
| Fires at 7,861 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to evidence or proof in various contexts""" |
| |
| fire = { |
| |
| ' evidence', ' Evidence', 'Evidence', 'evidence', ' evid', |
| |
| ' proof', |
| |
| ' indications', |
| |
| ' empirical', ' empir', |
| |
| ' evidenced', |
| |
| ' testim', |
| |
| ' Scripture', |
| |
| '________________________________', |
| |
| ' circumst', |
| |
| ' testimonies', |
| |
| ' conclusive', |
| |
| ' indis', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'evid-family', 1 |
|
|
| def prog_L06F21910(input_tokens): |
| """44 token families, 53 surface forms (train F1 of the uncapped optimum: 0.791). |
| Fires at 7,634 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of a specific name or title""" |
| |
| fire = { |
| |
| ' Motor', 'Motor', |
| |
| 'nar', ' Nar', 'Nar', ' nar', |
| |
| ' Rodrigo', |
| |
| ' Naval', |
| |
| ' Nissan', |
| |
| ' Lar', ' lar', |
| |
| ' Nak', |
| |
| 'agar', |
| |
| ' Viktor', |
| |
| ' SWAT', |
| |
| ' Arri', |
| |
| ' Atl', 'Atl', |
| |
| ' Inspector', |
| |
| ' Hyder', |
| |
| 'aru', |
| |
| ' Byrne', |
| |
| ' Narc', ' narc', |
| |
| ' aur', ' Aur', |
| |
| ' Hak', |
| |
| ' reimburse', ' reimbursement', |
| |
| ' Hannibal', |
| |
| ' criminally', |
| |
| ' Kak', |
| |
| ' Inner', |
| |
| ' Andromeda', |
| |
| ' Metall', |
| |
| ' Outer', |
| |
| ' Mé', |
| |
| ' Subaru', |
| |
| ' Naruto', |
| |
| ' Ley', |
| |
| 'orr', |
| |
| ' Saras', |
| |
| ' stimul', |
| |
| ' Nora', |
| |
| ' sher', |
| |
| 'ujah', |
| |
| ' Lup', |
| |
| ' Alph', |
| |
| 'iban', |
| |
| ' reimb', |
| |
| ' REM', |
| |
| ' narration', |
| |
| ' hypertension', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'motor-family', 1 |
|
|
| def prog_L06F21911(input_tokens): |
| """3 token families, 10 surface forms (train F1 of the uncapped optimum: 0.232). |
| Fires at 8,379 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "who" indicating questions of identity or responsibility""" |
| |
| fire = { |
| |
| ' who', ' Who', 'Who', 'who', ' WHO', 'WHO', |
| |
| ' whoever', ' Whoever', 'Whoever', |
| |
| 'omever', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'who-family', 1 |
|
|
| def prog_L06F21912(input_tokens): |
| """4 token families, 7 surface forms (train F1 of the uncapped optimum: 0.089). |
| Fires at 168,708 of 36,786,566 positions. |
| Fires at 0.46% 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: prominent political or social issues and activism-related terms""" |
| |
| fire = { |
| |
| "'", " '", |
| |
| ':', |
| |
| ' Trump', 'trump', ' TRUMP', |
| |
| 'We', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return "'-family", 1 |
|
|
| def prog_L06F21913(input_tokens): |
| """26 token families, 32 surface forms (train F1 of the uncapped optimum: 0.658). |
| Fires at 5,105 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the phrase "READ MORE." """ |
| |
| fire = { |
| |
| 'READ', ' READ', |
| |
| 'MAN', |
| |
| 'OW', |
| |
| 'GET', ' GET', |
| |
| 'WATCH', ' WATCH', |
| |
| 'rawdownloadcloneembedreportprint', |
| |
| 'EDIT', ' EDIT', |
| |
| 'SHARE', |
| |
| 'SEE', |
| |
| 'UAL', |
| |
| 'NEWS', |
| |
| 'POL', |
| |
| ' CET', |
| |
| 'LIST', |
| |
| 'JUST', |
| |
| 'PET', |
| |
| 'SUP', |
| |
| 'LLOW', |
| |
| 'PRE', |
| |
| 'ULL', |
| |
| '►', |
| |
| 'NOW', |
| |
| 'REP', 'reply', |
| |
| 'SIGN', |
| |
| 'RAW', |
| |
| ' FIX', 'FIX', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'read-family', 1 |
|
|
| def prog_L06F21914(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.166). |
| Fires at 134,711 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences 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_L06F21915(input_tokens): |
| """32 token families, 33 surface forms (train F1 of the uncapped optimum: 0.791). |
| Fires at 6,507 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the name "Austin" in various contexts""" |
| |
| fire = { |
| |
| ' Austin', 'Austin', |
| |
| ' Antonio', |
| |
| ' Banks', |
| |
| ' Sheffield', |
| |
| ' UT', |
| |
| ' Bristol', |
| |
| ' Cain', |
| |
| ' Barton', |
| |
| 'xton', |
| |
| ' Williamson', |
| |
| ' SX', |
| |
| ' Cork', |
| |
| ' CBD', |
| |
| ' Bangalore', |
| |
| ' PAX', |
| |
| ' Becker', |
| |
| ' Amar', |
| |
| ' Burgess', |
| |
| ' Cedar', |
| |
| ' Ashton', |
| |
| ' oak', |
| |
| ' Cary', |
| |
| ' Ci', |
| |
| ' Vaughan', |
| |
| 'aign', |
| |
| ' Augusta', |
| |
| 'dain', |
| |
| 'ageddon', |
| |
| ' aust', |
| |
| ' Avalon', |
| |
| ' ACS', |
| |
| ' avocado', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'austin-family', 1 |
|
|
| def prog_L06F21916(input_tokens): |
| """40 token families, 45 surface forms (train F1 of the uncapped optimum: 0.639). |
| Fires at 7,372 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the Eiffel Tower""" |
| |
| fire = { |
| |
| 'iff', 'iffs', 'IFF', |
| |
| 'Few', 'few', |
| |
| ' Cliff', |
| |
| ' sniff', |
| |
| ' Sec', |
| |
| ' Judy', |
| |
| ' Britt', |
| |
| ' Tiffany', |
| |
| 'rict', |
| |
| ' Forty', |
| |
| 'icket', |
| |
| ' tariff', |
| |
| ' Mob', |
| |
| ' Schiff', |
| |
| 'iem', |
| |
| 'iffin', |
| |
| ' Ticket', |
| |
| ' Piper', |
| |
| ' supp', |
| |
| 'arge', |
| |
| ' riff', |
| |
| 'iola', |
| |
| ' Thirty', |
| |
| 'ruff', |
| |
| 'atel', |
| |
| 'Diff', ' Diff', |
| |
| ' Clifford', |
| |
| 'ipt', |
| |
| ' MF', 'MF', |
| |
| 'Bey', |
| |
| ' Griff', |
| |
| ' prejud', |
| |
| 'iffe', |
| |
| 'ishop', |
| |
| 'igger', |
| |
| ' mism', |
| |
| 'aund', |
| |
| 'Font', |
| |
| ' byte', |
| |
| ' Tiff', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'iff-family', 1 |
|
|
| def prog_L06F21917(input_tokens): |
| """14 token families, 25 surface forms (train F1 of the uncapped optimum: 0.156). |
| Fires at 708,064 of 36,786,566 positions. |
| Fires at 1.92% 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 events involving conflict or violence""" |
| |
| fire = { |
| |
| ' a', |
| |
| ' has', |
| |
| ' by', |
| |
| ' after', |
| |
| ' been', |
| |
| ' week', ' weeks', 'week', |
| |
| ' which', |
| |
| ' month', ' months', 'month', ' Months', 'months', |
| |
| ' days', ' day', 'day', 'Days', |
| |
| ' recent', ' recently', ' Recently', |
| |
| 'In', |
| |
| ' year', |
| |
| ' according', |
| |
| ' since', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'a-family', 1 |
|
|
| def prog_L06F21918(input_tokens): |
| """19 token families, 48 surface forms (train F1 of the uncapped optimum: 0.213). |
| Fires at 177,013 of 36,786,566 positions. |
| Fires at 0.48% 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 reports and analyses""" |
| |
| fire = { |
| |
| ' report', ' reports', ' reported', ' reporter', ' Report', ' reporting', ' Reports', ' REPORT', |
| |
| ' rules', ' ruling', ' ruled', ' Rules', ' rulings', |
| |
| ' research', ' researchers', ' Research', ' researcher', |
| |
| ' data', |
| |
| ' law', |
| |
| ' poll', ' Poll', ' polling', |
| |
| ' decision', ' Decision', |
| |
| ' investigation', |
| |
| ' meeting', ' meetings', ' Meeting', |
| |
| ' plan', |
| |
| ' policy', |
| |
| ' funding', ' funds', ' Funding', ' Funds', |
| |
| ' documents', ' document', ' Documents', |
| |
| ' website', |
| |
| ' analysis', ' Analysis', 'analysis', |
| |
| ' response', ' Response', |
| |
| ' efforts', |
| |
| ' coverage', ' Coverage', |
| |
| ' survey', ' surveys', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'report-family', 1 |
|
|
| def prog_L06F21919(input_tokens): |
| """38 token families, 64 surface forms (train F1 of the uncapped optimum: 0.408). |
| Fires at 104,899 of 36,786,566 positions.""" |
| |
| """prev SAE description: gerunds and present participles that emphasize ongoing actions or processes""" |
| |
| fire = { |
| |
| ' making', 'Making', ' Making', |
| |
| ' being', |
| |
| ' getting', 'Getting', ' Getting', 'getting', |
| |
| ' creating', ' Creating', 'Creating', |
| |
| ' having', |
| |
| ' taking', |
| |
| ' playing', ' Playing', |
| |
| ' building', 'Building', |
| |
| ' putting', 'Putting', ' Putting', |
| |
| ' helping', |
| |
| ' providing', |
| |
| ' keeping', 'Keeping', ' Keeping', |
| |
| ' giving', 'Giving', ' Giving', |
| |
| ' bringing', ' Bringing', |
| |
| ' finding', ' Finding', 'finding', |
| |
| ' developing', |
| |
| ' writing', |
| |
| ' winning', |
| |
| ' seeing', ' Seeing', 'Seeing', 'seeing', |
| |
| ' selling', ' Selling', |
| |
| ' killing', |
| |
| ' paying', |
| |
| ' protecting', |
| |
| ' supporting', ' Supporting', |
| |
| ' losing', |
| |
| ' improving', |
| |
| ' raising', |
| |
| ' buying', |
| |
| ' reducing', |
| |
| ' passing', ' Passing', ' Passed', |
| |
| ' entering', |
| |
| ' sending', ' Sending', |
| |
| ' promoting', |
| |
| ' failing', |
| |
| ' dealing', |
| |
| ' producing', |
| |
| ' removing', |
| |
| ' reaching', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'mak-family', 1 |
|
|
| def prog_L06F21920(input_tokens): |
| """6 token families, 10 surface forms (train F1 of the uncapped optimum: 0.754). |
| Fires at 6,832 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the word "Me." """ |
| |
| fire = { |
| |
| ' Me', 'me', 'Me', ' ME', 'ME', |
| |
| ' Megan', |
| |
| ' Meh', |
| |
| ' measles', |
| |
| ' Moe', |
| |
| ' Mei', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'me-family', 1 |
|
|
| def prog_L06F21921(input_tokens): |
| """2 token families, 7 surface forms (train F1 of the uncapped optimum: 0.067). |
| Fires at 33,260 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases expressing intentions or desires to take action""" |
| |
| fire = { |
| |
| ' to', 'to', ' To', ' TO', |
| |
| ' should', 'should', ' SHOULD', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'to-family', 1 |
|
|
| def prog_L06F21922(input_tokens): |
| """36 token families, 44 surface forms (train F1 of the uncapped optimum: 0.650). |
| Fires at 8,645 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "poster" and its variations""" |
| |
| fire = { |
| |
| ' poster', ' posters', ' Poster', |
| |
| 'oster', |
| |
| ' billboard', ' billboards', |
| |
| ' mural', ' mur', |
| |
| ' Pul', |
| |
| ' flyers', ' flyer', |
| |
| ' framed', |
| |
| ' commenters', ' commenter', |
| |
| ' sticker', |
| |
| ' Paramount', |
| |
| 'boards', |
| |
| ' infographic', |
| |
| ' posing', |
| |
| 'paper', 'papers', |
| |
| ' vide', ' Vide', |
| |
| ' bumper', |
| |
| 'aper', |
| |
| ' presentations', |
| |
| 'pal', |
| |
| ' bulletin', |
| |
| ' pics', |
| |
| ' prints', |
| |
| 'pad', |
| |
| ' projector', |
| |
| ' PCB', |
| |
| ' wallpaper', |
| |
| ' sten', |
| |
| ' resur', |
| |
| ' crucifix', |
| |
| ' plaster', |
| |
| 'pter', |
| |
| ' PowerPoint', |
| |
| ' posterior', |
| |
| ' pall', |
| |
| ' Thumbnails', |
| |
| ' Moder', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'post-family', 1 |
|
|
| def prog_L06F21923(input_tokens): |
| """12 token families, 17 surface forms (train F1 of the uncapped optimum: 0.384). |
| Fires at 8,338 of 36,786,566 positions.""" |
| |
| """prev SAE description: concepts and discussions centered around significance or importance""" |
| |
| fire = { |
| |
| ' matter', ' matters', ' Matter', ' Matters', 'matter', |
| |
| ' applies', |
| |
| ' looms', |
| |
| ' cares', |
| |
| ' sucks', |
| |
| ' mattered', |
| |
| ' rang', |
| |
| ' MAT', |
| |
| ' suffice', |
| |
| ' palpable', |
| |
| ' paramount', |
| |
| 'tains', 'tained', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'matt-family', 1 |
|
|
| def prog_L06F21924(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.264). |
| Fires at 25,985 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "described" and its associated phrases""" |
| |
| fire = { |
| |
| ' as', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'as-family', 1 |
|
|
| def prog_L06F21925(input_tokens): |
| """2 token families, 3 surface forms (train F1 of the uncapped optimum: 0.639). |
| Fires at 9,841 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to links or connections in various contexts""" |
| |
| fire = { |
| |
| ' between', ' Between', |
| |
| 'pmwiki', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'between-family', 1 |
|
|
| def prog_L06F21926(input_tokens): |
| """44 token families, 52 surface forms (train F1 of the uncapped optimum: 0.555). |
| Fires at 8,340 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to confessions and admission of guilt""" |
| |
| fire = { |
| |
| ' confess', ' confession', |
| |
| ' pact', ' Pact', |
| |
| ' Shawn', |
| |
| ' confessed', |
| |
| ' Reve', 'reve', |
| |
| ' Sud', |
| |
| ' Rudy', |
| |
| ' candid', |
| |
| 'ession', 'essional', |
| |
| ' contr', |
| |
| ' admitting', |
| |
| ' betray', ' betrayed', |
| |
| ' bragging', |
| |
| ' secession', |
| |
| ' RAW', |
| |
| ' volunteered', |
| |
| ' sins', |
| |
| ' profess', |
| |
| ' baptized', ' Bapt', ' bapt', |
| |
| ' confessions', |
| |
| ' Passion', |
| |
| ' Calvin', |
| |
| ' crocod', |
| |
| ' Tid', |
| |
| ' declarations', |
| |
| ' boasting', |
| |
| ' remorse', |
| |
| ' Couch', |
| |
| ' Nath', |
| |
| ' Dah', |
| |
| 'akable', |
| |
| ' baptism', |
| |
| ' blush', |
| |
| ' incrim', |
| |
| 'chenko', |
| |
| ' Freud', |
| |
| ' cassette', |
| |
| ' cyan', |
| |
| 'itent', |
| |
| ' autobi', |
| |
| ' weep', |
| |
| ' repent', ' repentance', |
| |
| 'conv', |
| |
| ' trembling', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'confes-family', 1 |
|
|
| def prog_L06F21927(input_tokens): |
| """4 token families, 10 surface forms (train F1 of the uncapped optimum: 0.801). |
| Fires at 12,345 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to various medical or training centers""" |
| |
| fire = { |
| |
| ' Center', ' center', ' centers', 'Center', 'center', ' CENT', |
| |
| ' Centre', ' centre', |
| |
| ' centres', |
| |
| ' Appeals', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'cent-family', 1 |
|
|
| def prog_L06F21928(input_tokens): |
| """5 token families, 6 surface forms (train F1 of the uncapped optimum: 0.272). |
| Fires at 7,791 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the phrase "break down." """ |
| |
| fire = { |
| |
| ' down', 'down', |
| |
| ' breakdown', |
| |
| 'lated', |
| |
| 'border', |
| |
| 'struct', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'down-family', 1 |
|
|
| def prog_L06F21929(input_tokens): |
| """41 token families, 42 surface forms (train F1 of the uncapped optimum: 0.239). |
| Fires at 204,601 of 36,786,566 positions. |
| Fires at 0.56% 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 dates and times in the text""" |
| |
| fire = { |
| |
| ' year', |
| |
| ' last', |
| |
| ' week', |
| |
| ' ago', |
| |
| ' Monday', |
| |
| ' Tuesday', |
| |
| ' Wednesday', |
| |
| ' Friday', |
| |
| ' Thursday', |
| |
| ' month', |
| |
| ' night', |
| |
| ' Sunday', |
| |
| ' Saturday', |
| |
| ' March', |
| |
| ' May', |
| |
| ' 2015', |
| |
| ' morning', |
| |
| ' late', ' lately', |
| |
| ' July', |
| |
| ' June', |
| |
| ' 2014', |
| |
| ' 2012', |
| |
| ' 2013', |
| |
| ' September', |
| |
| ' January', |
| |
| ' April', |
| |
| ' October', |
| |
| ' August', |
| |
| ' 2011', |
| |
| ' November', |
| |
| ' earlier', |
| |
| ' weekend', |
| |
| ' afternoon', |
| |
| ' 2010', |
| |
| ' December', |
| |
| ' February', |
| |
| ' 2008', |
| |
| ' summer', |
| |
| ' 2009', |
| |
| ' yesterday', |
| |
| ' 2007', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'year-family', 1 |
|
|
| def prog_L06F21930(input_tokens): |
| """6 token families, 6 surface forms (train F1 of the uncapped optimum: 0.158). |
| Fires at 99,073 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating requests or actions taken by groups of people""" |
| |
| fire = { |
| |
| ' are', |
| |
| ' have', |
| |
| ' were', |
| |
| ' often', |
| |
| ' aren', |
| |
| ' seem', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'are-family', 1 |
|
|
| def prog_L06F21931(input_tokens): |
| """7 token families, 12 surface forms (train F1 of the uncapped optimum: 0.070). |
| Fires at 159,610 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: phrases related to action-oriented themes and activities""" |
| |
| fire = { |
| |
| ',', |
| |
| ' (', |
| |
| ' against', ' Against', |
| |
| 'CLOSE', |
| |
| ' between', |
| |
| ' action', ' actions', 'actions', |
| |
| ' involving', ' involved', ' involves', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F21932(input_tokens): |
| """29 token families, 31 surface forms (train F1 of the uncapped optimum: 0.806). |
| Fires at 4,326 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the prefix "com" in various contexts""" |
| |
| fire = { |
| |
| ' com', ' Com', ' COM', |
| |
| ' comm', |
| |
| ' comet', |
| |
| ' coma', |
| |
| ' gru', |
| |
| ' android', |
| |
| ' org', |
| |
| ' comple', |
| |
| ' emer', |
| |
| ' Carmen', |
| |
| ' incon', |
| |
| ' benign', |
| |
| ' java', |
| |
| ' consum', |
| |
| ' uncom', |
| |
| ' ris', |
| |
| ' congen', |
| |
| ' rocking', |
| |
| ' prod', |
| |
| ' sust', |
| |
| ' accus', |
| |
| ' indis', |
| |
| ' incom', |
| |
| 'י�', |
| |
| ' brun', |
| |
| ' produ', |
| |
| ' evid', |
| |
| ' unimagin', |
| |
| ' educ', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'com-family', 1 |
|
|
| def prog_L06F21933(input_tokens): |
| """49 token families, 52 surface forms (train F1 of the uncapped optimum: 0.687). |
| Fires at 11,481 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of memorials or commemorative events""" |
| |
| fire = { |
| |
| ' Memorial', ' memorial', |
| |
| ' Independence', |
| |
| ' Cyber', |
| |
| ' Veterans', |
| |
| ' Holiday', |
| |
| ' Trophy', |
| |
| ' lobby', |
| |
| ' Civic', |
| |
| ' Monument', |
| |
| 'Mem', ' MEM', |
| |
| ' Parkway', |
| |
| 'EMBER', |
| |
| ' Coliseum', |
| |
| ' Clara', |
| |
| ' Jurassic', |
| |
| ' Highland', ' Highlands', |
| |
| ' Huntington', |
| |
| ' Wembley', |
| |
| ' Governors', |
| |
| ' Harbor', |
| |
| ' Hampton', |
| |
| ' Presidents', |
| |
| ' Accountability', |
| |
| ' bumper', |
| |
| ' Rockefeller', |
| |
| ' Hancock', |
| |
| ' Glacier', |
| |
| ' Harbour', |
| |
| ' Maritime', |
| |
| ' Hammer', |
| |
| ' Friendship', |
| |
| 'ummer', |
| |
| ' Buckingham', |
| |
| ' Remem', |
| |
| ' Lunar', |
| |
| ' Statue', |
| |
| ' MGM', |
| |
| ' RSA', |
| |
| ' Clapper', |
| |
| ' Lucky', |
| |
| ' Hamp', |
| |
| ' Missile', |
| |
| ' Cyp', |
| |
| ' Lawn', |
| |
| 'umann', |
| |
| ' Countdown', |
| |
| ' Neurolog', |
| |
| ' Lutheran', |
| |
| ' Lazarus', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'memor-family', 1 |
|
|
| def prog_L06F21934(input_tokens): |
| """39 token families, 48 surface forms (train F1 of the uncapped optimum: 0.714). |
| Fires at 5,511 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "gro" indicating themes of growth or collectivism""" |
| |
| fire = { |
| |
| ' Gro', ' gro', 'gro', 'Gro', |
| |
| 'agan', |
| |
| ' Grove', |
| |
| ' Fro', ' fro', |
| |
| 'thro', ' thro', ' Thro', |
| |
| ' groom', ' grooming', |
| |
| 'isco', |
| |
| 'FO', |
| |
| ' Gri', ' gri', |
| |
| 'ockey', |
| |
| ' groin', |
| |
| ' blo', |
| |
| ' heel', |
| |
| 'GI', |
| |
| ' gymn', |
| |
| ' headaches', |
| |
| ' flo', |
| |
| ' shri', |
| |
| ' lou', |
| |
| 'Gra', |
| |
| ' gasp', ' gasped', |
| |
| ' hob', |
| |
| ' funk', |
| |
| ' agon', |
| |
| ' groove', |
| |
| ' GH', |
| |
| ' awkwardly', |
| |
| ' grun', |
| |
| ' grape', |
| |
| 'JO', |
| |
| ' atro', |
| |
| ' invol', |
| |
| ' stag', |
| |
| ' grou', |
| |
| ' impro', |
| |
| ' grunt', |
| |
| ' flatt', |
| |
| 'gre', |
| |
| 'eals', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'gro-family', 1 |
|
|
| def prog_L06F21935(input_tokens): |
| """24 token families, 93 surface forms (train F1 of the uncapped optimum: 0.289). |
| Fires at 161,736 of 36,786,566 positions. |
| Fires at 0.44% 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 hardback and limited edition printed materials""" |
| |
| fire = { |
| |
| ' book', ' books', 'book', 'books', 'Books', 'Book', 'BOOK', |
| |
| ' edition', ' Edition', 'edition', ' EDITION', |
| |
| ' series', 'series', ' serial', |
| |
| ' release', 'release', 'Release', |
| |
| ' version', |
| |
| ' collection', ' Collection', 'collection', 'Collection', |
| |
| ' short', ' shorts', |
| |
| ' pack', ' Pack', ' packs', ' Packs', 'pack', 'packs', 'Pack', ' PACK', |
| |
| ' album', ' albums', 'album', |
| |
| ' original', ' Original', 'original', |
| |
| ' cover', ' covers', 'cover', ' Cover', 'Cover', ' COVER', |
| |
| ' print', ' printed', ' prints', 'print', 'prints', 'printed', ' Printed', |
| |
| ' limited', ' Limited', 'Limited', 'limited', ' LIMITED', |
| |
| ' digital', ' digit', 'Digital', 'digital', |
| |
| ' novel', ' novels', ' Novel', |
| |
| ' classic', ' Classic', 'Classic', 'classic', |
| |
| ' disc', ' Disc', 'disc', ' discs', 'Disc', |
| |
| ' comic', |
| |
| ' sets', ' Set', ' Sets', 'sets', |
| |
| ' releases', 'released', ' Releases', |
| |
| ' download', ' downloadable', ' downloads', ' Downloads', |
| |
| ' volume', ' Volume', 'Volume', 'volume', |
| |
| ' DVD', 'DVD', ' DVDs', |
| |
| ' featuring', ' Featuring', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'book-family', 1 |
|
|
| def prog_L06F21936(input_tokens): |
| """40 token families, 57 surface forms (train F1 of the uncapped optimum: 0.364). |
| Fires at 62,793 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to specific sports organizations and competitions""" |
| |
| fire = { |
| |
| ' League', 'League', 'league', |
| |
| 'L', |
| |
| ' NBA', 'NBA', |
| |
| ' FC', 'FC', |
| |
| ' Cup', 'cup', |
| |
| ' Soccer', |
| |
| ' Championship', |
| |
| ' MLS', |
| |
| 'BA', |
| |
| ' Rugby', ' rugby', |
| |
| ' basketball', ' Basketball', |
| |
| ' Series', 'Series', |
| |
| ' Hockey', |
| |
| ' Test', 'Test', |
| |
| 'port', 'ports', |
| |
| ' Open', |
| |
| ' AFL', |
| |
| ' SEC', 'SEC', |
| |
| ' NHL', |
| |
| 'HL', |
| |
| 'BC', |
| |
| 'AC', |
| |
| ' Tour', 'Tour', ' TOUR', |
| |
| ' NRL', |
| |
| 'FL', |
| |
| 'GP', ' GP', |
| |
| 'GA', |
| |
| 'CS', |
| |
| 'weight', |
| |
| ' Universe', |
| |
| ' Ten', |
| |
| 'ollywood', |
| |
| 'MA', |
| |
| ' GT', |
| |
| 'BO', |
| |
| 'Sports', 'sports', ' SPORTS', |
| |
| 'FA', |
| |
| ' ACC', 'ACC', |
| |
| ' Racing', |
| |
| ' Baseball', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'league-family', 1 |
|
|
| def prog_L06F21937(input_tokens): |
| """1 token families, 3 surface forms (train F1 of the uncapped optimum: 0.326). |
| Fires at 59,090 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases that start with "one thing that" to identify key points or themes""" |
| |
| fire = { |
| |
| ' that', ' THAT', ' thats', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'that-family', 1 |
|
|
| def prog_L06F21938(input_tokens): |
| """2 token families, 4 surface forms (train F1 of the uncapped optimum: 0.490). |
| Fires at 15,994 of 36,786,566 positions.""" |
| |
| """prev SAE description: the phrase "at the end of the day." """ |
| |
| fire = { |
| |
| ' day', 'Day', ' DAY', |
| |
| 'pt', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'day-family', 1 |
|
|
| def prog_L06F21939(input_tokens): |
| """49 token families, 64 surface forms (train F1 of the uncapped optimum: 0.707). |
| Fires at 18,358 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the name "Crow." """ |
| |
| fire = { |
| |
| ' Hawks', ' hawk', ' Hawk', |
| |
| ' Crow', ' crow', 'crow', |
| |
| ' Buck', |
| |
| ' LGBTQ', |
| |
| ' Jew', 'Jew', |
| |
| ' Brow', ' brow', 'Brow', 'brow', |
| |
| ' Chrom', ' chrom', |
| |
| ' Counsel', |
| |
| ' chains', ' Chains', |
| |
| ' Bowie', |
| |
| ' Schneider', |
| |
| ' divisive', |
| |
| 'kes', |
| |
| ' Scots', |
| |
| ' bees', ' Bees', |
| |
| ' Gaz', |
| |
| ' mourn', |
| |
| ' slur', ' slurs', |
| |
| ' counselor', ' counselors', |
| |
| ' Ace', |
| |
| ' queer', |
| |
| ' Cage', |
| |
| ' claw', ' Claw', |
| |
| ' Jung', |
| |
| ' Cran', ' cran', |
| |
| ' Kang', |
| |
| ' Cobb', |
| |
| ' Swan', |
| |
| ' Crowley', |
| |
| ' scrub', |
| |
| ' Journey', |
| |
| ' blues', |
| |
| ' Pike', |
| |
| ' Roku', |
| |
| ' Kem', |
| |
| ' Shel', |
| |
| ' Cay', |
| |
| ' dwell', |
| |
| ' Slack', |
| |
| ' Cock', |
| |
| ' Bun', |
| |
| ' segregation', |
| |
| ' Daniels', |
| |
| ' discriminatory', |
| |
| ' Nun', |
| |
| ' wrest', |
| |
| ' Norse', |
| |
| ' Bones', |
| |
| ' Cary', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'hawk-family', 1 |
|
|
| def prog_L06F21940(input_tokens): |
| """19 token families, 23 surface forms (train F1 of the uncapped optimum: 0.370). |
| Fires at 69,559 of 36,786,566 positions.""" |
| |
| """prev SAE description: names of football players and related statistics""" |
| |
| fire = { |
| |
| ' City', |
| |
| ' scored', ' scoring', ' scorer', |
| |
| ' striker', ' strikers', |
| |
| ' Mourinho', |
| |
| 'ane', |
| |
| ' midfielder', ' midfield', |
| |
| ' Rodgers', |
| |
| 'ini', |
| |
| 'ovic', |
| |
| 'ba', |
| |
| ' Silva', |
| |
| ' Wenger', |
| |
| ' Gaal', |
| |
| ' Klopp', |
| |
| ' Ronaldo', |
| |
| ' Rooney', |
| |
| ' Kane', |
| |
| ' Messi', |
| |
| ' Brady', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'city-family', 1 |
|
|
| def prog_L06F21941(input_tokens): |
| """44 token families, 53 surface forms (train F1 of the uncapped optimum: 0.558). |
| Fires at 12,659 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of transcripts""" |
| |
| fire = { |
| |
| ' transcript', ' Transcript', ' transcripts', |
| |
| ' audio', 'Audio', |
| |
| ' text', |
| |
| ' excerpt', ' excerpts', |
| |
| 'cript', |
| |
| ' Minutes', |
| |
| ' Copy', |
| |
| ' recordings', |
| |
| ' transc', |
| |
| ' tapes', |
| |
| ' replay', |
| |
| ' entirety', |
| |
| ' Proceedings', |
| |
| ' purse', |
| |
| ' Rubin', |
| |
| ' snippet', ' snippets', |
| |
| ' Chomsky', |
| |
| ' clips', |
| |
| 'cerpt', |
| |
| 'Screenshot', |
| |
| 'Brend', |
| |
| ' dialect', |
| |
| 'acket', 'ackets', |
| |
| 'ript', 'RIPT', |
| |
| ' aloud', |
| |
| ' overheard', |
| |
| ' syll', |
| |
| ' receipts', |
| |
| ' redacted', |
| |
| ' Qur', |
| |
| ' VIDEO', |
| |
| ' subtitles', |
| |
| ' paraph', |
| |
| ' citation', ' Citation', |
| |
| 'reen', |
| |
| ' transcription', |
| |
| ' narrator', |
| |
| ' corpus', ' Corpus', |
| |
| ' Brenda', |
| |
| ' subtitle', |
| |
| ' condensed', |
| |
| ' citations', |
| |
| ' PLoS', |
| |
| ' phon', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'transcript-family', 1 |
|
|
| def prog_L06F21942(input_tokens): |
| """34 token families, 35 surface forms (train F1 of the uncapped optimum: 0.287). |
| Fires at 36,083 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to various government departments and their functions""" |
| |
| fire = { |
| |
| ' Health', |
| |
| ' Justice', |
| |
| ' Human', |
| |
| ' Homeland', |
| |
| ' Public', |
| |
| ' Control', |
| |
| ' Affairs', |
| |
| ' Education', |
| |
| ' Transportation', |
| |
| ' Energy', |
| |
| ' Immigration', |
| |
| ' Foreign', |
| |
| ' Environmental', |
| |
| ' Disease', |
| |
| ' Labor', |
| |
| ' Economic', |
| |
| ' Environment', 'Environment', |
| |
| ' Revenue', |
| |
| ' Agriculture', |
| |
| ' Trade', |
| |
| ' Customs', |
| |
| ' Appeals', |
| |
| ' Management', |
| |
| ' Natural', |
| |
| ' Statistics', |
| |
| ' Investigation', |
| |
| ' Resources', |
| |
| ' Rex', |
| |
| ' Border', |
| |
| ' Medicine', |
| |
| ' Commerce', |
| |
| ' Veterans', |
| |
| ' Interior', |
| |
| ' Housing', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'health-family', 1 |
|
|
| def prog_L06F21943(input_tokens): |
| """7 token families, 22 surface forms (train F1 of the uncapped optimum: 0.119). |
| Fires at 615,406 of 36,786,566 positions. |
| Fires at 1.67% 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: technical terms and concepts relevant to programming and technology""" |
| |
| fire = { |
| |
| '�', ' �', |
| |
| ')', |
| |
| 'm', 'M', ' M', ' m', |
| |
| 'C', ' C', 'c', ' c', |
| |
| ' A', 'a', 'ía', |
| |
| ' B', 'B', 'b', ' b', |
| |
| 'D', 'd', ' d', ' dé', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '�-family', 1 |
|
|
| def prog_L06F21944(input_tokens): |
| """7 token families, 22 surface forms (train F1 of the uncapped optimum: 0.161). |
| Fires at 518,135 of 36,786,566 positions. |
| Fires at 1.41% 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 formal documents and structures""" |
| |
| fire = { |
| |
| '-', |
| |
| ' new', ' news', ' newly', 'new', ' newer', |
| |
| 'year', ' yearly', |
| |
| 'old', ' old', 'olding', |
| |
| ' high', ' higher', ' highly', 'high', 'highly', |
| |
| ' statement', 'state', 'statement', |
| |
| ' long', 'long', ' LONG', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '--family', 1 |
|
|
| def prog_L06F21945(input_tokens): |
| """47 token families, 62 surface forms (train F1 of the uncapped optimum: 0.689). |
| Fires at 24,865 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to backpacks and their related contexts""" |
| |
| fire = { |
| |
| 'pack', 'packs', 'packing', ' Packs', |
| |
| ' hiking', ' hikes', ' hikers', |
| |
| ' hike', |
| |
| '?!', |
| |
| ' monk', ' monks', ' Monk', |
| |
| 'án', |
| |
| ' refuge', ' Refuge', ' Refugees', |
| |
| ' Tra', |
| |
| ' Kaepernick', |
| |
| 'itive', |
| |
| ' HQ', 'HQ', |
| |
| 'TF', ' TF', 'tf', |
| |
| ' backpack', |
| |
| 'iman', |
| |
| 'itarian', |
| |
| 'Wal', |
| |
| ' Hamburg', |
| |
| ' Stefan', |
| |
| 'rio', |
| |
| 'ban', |
| |
| 'ulk', |
| |
| ' squat', ' squats', |
| |
| ' Ord', |
| |
| ' Hut', ' hut', |
| |
| ' Hulk', |
| |
| ' Riot', |
| |
| ' Jeep', |
| |
| ' Bundy', |
| |
| ' undrafted', |
| |
| ' Ricardo', |
| |
| ' Bund', |
| |
| ' summ', ' summed', |
| |
| ' Dunn', |
| |
| ' fugitive', |
| |
| ' Jorge', |
| |
| ' Johnston', |
| |
| 'uchin', |
| |
| 'ampa', |
| |
| 'stad', |
| |
| ' detain', |
| |
| 'jiang', |
| |
| ' doub', |
| |
| ' Huff', |
| |
| 'tall', |
| |
| 'stadt', |
| |
| 'ouf', |
| |
| ' fasting', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'pack-family', 1 |
|
|
| def prog_L06F21946(input_tokens): |
| """39 token families, 56 surface forms (train F1 of the uncapped optimum: 0.635). |
| Fires at 19,254 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "praise" and its variations""" |
| |
| fire = { |
| |
| ' celebrated', ' celebrating', ' celebrates', |
| |
| ' criticized', ' criticised', |
| |
| ' praised', ' Pra', |
| |
| ' praise', |
| |
| ' defended', ' defends', |
| |
| ' hailed', |
| |
| ' touted', ' touting', |
| |
| ' welcomed', |
| |
| ' endorsed', ' endorsing', |
| |
| ' criticizing', |
| |
| ' thanked', ' thanking', |
| |
| ' mocked', ' mocking', |
| |
| ' credited', |
| |
| ' praising', ' praises', |
| |
| ' applaud', ' applauded', |
| |
| ' lauded', ' laud', |
| |
| ' tribute', |
| |
| ' honored', ' honoring', |
| |
| ' cheered', ' cheering', |
| |
| ' criticize', |
| |
| ' compliment', ' compliments', |
| |
| ' approving', ' approves', |
| |
| ' glowing', |
| |
| ' rewarded', |
| |
| ' admire', |
| |
| ' admired', ' admirable', |
| |
| ' condemning', |
| |
| ' salute', |
| |
| ' accol', |
| |
| ' honorary', |
| |
| ' glor', |
| |
| ' admiration', |
| |
| ' congratulated', |
| |
| ' retweet', |
| |
| ' lamented', |
| |
| ' congratulate', |
| |
| ' congrat', |
| |
| ' complimentary', |
| |
| ' commend', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'celebrat-family', 1 |
|
|
| def prog_L06F21947(input_tokens): |
| """37 token families, 47 surface forms (train F1 of the uncapped optimum: 0.676). |
| Fires at 12,629 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to bicycles and biking""" |
| |
| fire = { |
| |
| ' bike', ' Bike', 'bike', |
| |
| ' bicycle', ' Bicycle', |
| |
| ' bikes', ' biking', |
| |
| ' cycling', 'cycl', 'cycles', |
| |
| ' motorcycle', |
| |
| ' cyclists', |
| |
| ' bicycles', ' bicycl', |
| |
| ' wheels', |
| |
| 'ikes', |
| |
| ' Patel', |
| |
| ' pistol', |
| |
| 'cycle', ' Cycle', |
| |
| ' Harley', |
| |
| ' puppy', |
| |
| ' cruiser', ' Cruiser', |
| |
| ' motorcycles', |
| |
| ' pedal', |
| |
| 'ooter', 'ooters', |
| |
| ' soda', |
| |
| ' cigarette', |
| |
| ' BM', |
| |
| ' thumbs', |
| |
| ' Shim', |
| |
| 'ikers', |
| |
| ' goats', |
| |
| 'asaki', |
| |
| ' bip', |
| |
| ' BI', |
| |
| ' stationary', |
| |
| ' bicy', |
| |
| ' cane', |
| |
| ' tractor', |
| |
| 'icycle', |
| |
| ' pedals', |
| |
| ' cabbage', |
| |
| ' lightsaber', |
| |
| 'blade', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'bike-family', 1 |
|
|
| def prog_L06F21948(input_tokens): |
| """12 token families, 20 surface forms (train F1 of the uncapped optimum: 0.794). |
| Fires at 5,760 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "double" """ |
| |
| fire = { |
| |
| ' double', ' Double', 'double', 'Double', |
| |
| ' triple', ' Triple', |
| |
| ' doubled', ' doubles', ' doubling', |
| |
| ' Single', 'Single', 'single', |
| |
| ' doub', |
| |
| 'Cross', |
| |
| ' shortcut', |
| |
| ' brute', |
| |
| ' Shift', |
| |
| ' duplicate', |
| |
| 'BLE', |
| |
| 'ouble', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'double-family', 1 |
|
|
| def prog_L06F21949(input_tokens): |
| """3 token families, 10 surface forms (train F1 of the uncapped optimum: 0.723). |
| Fires at 26,798 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to individuals in leadership positions or titles""" |
| |
| fire = { |
| |
| ' President', ' president', 'President', 'president', ' presidents', ' Presidents', ' PRESIDENT', |
| |
| ' VP', 'VP', |
| |
| ' treasurer', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'president-family', 1 |
|
|
| def prog_L06F21950(input_tokens): |
| """12 token families, 21 surface forms (train F1 of the uncapped optimum: 0.068). |
| Fires at 233,260 of 36,786,566 positions. |
| Fires at 0.63% 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: occurrences of personal pronouns and indicators of collective experiences or actions""" |
| |
| fire = { |
| |
| 'The', |
| |
| ' new', 'NEW', ' NEW', |
| |
| ' said', |
| |
| ' say', ' says', |
| |
| 'A', |
| |
| ' can', ' Can', 'CAN', |
| |
| ' We', 'We', |
| |
| ' could', |
| |
| 'But', |
| |
| ' man', ' MAN', |
| |
| ' help', ' helped', ' helping', |
| |
| ' --', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'the-family', 1 |
|
|
| def prog_L06F21951(input_tokens): |
| """8 token families, 28 surface forms (train F1 of the uncapped optimum: 0.206). |
| Fires at 400,591 of 36,786,566 positions. |
| Fires at 1.09% 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 notable films and cultural icons, particularly in contexts involving tragedy or social commentary""" |
| |
| fire = { |
| |
| ' Trump', 'Trump', |
| |
| ' War', ' war', ' Wars', 'war', ' wars', 'Wars', ' WAR', |
| |
| ' Obama', |
| |
| ' car', ' cars', 'car', ' Cars', 'cars', |
| |
| ' school', ' schools', ' School', 'school', ' Schools', ' schooling', |
| |
| ' election', 'election', ' Election', |
| |
| ' world', |
| |
| ' children', ' Children', 'children', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'trump-family', 1 |
|
|
| def prog_L06F21952(input_tokens): |
| """5 token families, 5 surface forms (train F1 of the uncapped optimum: 0.486). |
| Fires at 15,699 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "I" and look for an empty document identifier""" |
| |
| fire = { |
| |
| 'I', |
| |
| 'II', |
| |
| ' IPA', |
| |
| 'AI', |
| |
| 'ARI', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'i-family', 1 |
|
|
| def prog_L06F21953(input_tokens): |
| """48 token families, 62 surface forms (train F1 of the uncapped optimum: 0.699). |
| Fires at 20,104 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "Has" along with variations of the term "bro." """ |
| |
| fire = { |
| |
| ' Has', 'has', 'Has', ' HAS', |
| |
| 'bro', |
| |
| ' Had', |
| |
| ' Junior', |
| |
| ' Expo', |
| |
| 'icon', 'icons', |
| |
| ' Tes', 'tes', 'Tes', |
| |
| 'ardo', |
| |
| ' crossover', |
| |
| 'tro', 'Tro', |
| |
| ' LEGO', |
| |
| ' CES', |
| |
| ' Transformers', 'Transform', |
| |
| ' Devin', |
| |
| ' Yosh', |
| |
| 'produ', ' Produ', |
| |
| 'ibo', |
| |
| ' Hath', ' hath', |
| |
| ' Eb', ' EB', |
| |
| ' Hodg', |
| |
| 'veh', 'Veh', ' veh', |
| |
| ' Mes', |
| |
| ' turbo', ' Turbo', |
| |
| ' Nunes', |
| |
| ' Jal', |
| |
| ' Eid', |
| |
| ' Harmon', |
| |
| 'hib', |
| |
| ' Toys', |
| |
| ' Abrams', |
| |
| ' defy', |
| |
| ' defied', |
| |
| 'comm', |
| |
| 'wayne', |
| |
| ' Applied', |
| |
| ' Xavier', |
| |
| ' Idlib', |
| |
| ' cis', |
| |
| ' Unc', |
| |
| ' escal', |
| |
| ' Bac', |
| |
| 'igen', |
| |
| 'endor', |
| |
| 'rex', |
| |
| ' Jav', |
| |
| ' Cir', |
| |
| ' Hobby', |
| |
| ' Gau', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'has-family', 1 |
|
|
| def prog_L06F21954(input_tokens): |
| """2 token families, 3 surface forms (train F1 of the uncapped optimum: 0.093). |
| Fires at 48,343 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to deaths or incidents involving individuals""" |
| |
| fire = { |
| |
| ' of', ' Of', |
| |
| ' killed', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'of-family', 1 |
|
|
| def prog_L06F21955(input_tokens): |
| """8 token families, 12 surface forms (train F1 of the uncapped optimum: 0.802). |
| Fires at 9,316 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of "Internet." """ |
| |
| fire = { |
| |
| ' Internet', ' internet', 'Internet', 'internet', |
| |
| ' Net', |
| |
| ' IoT', |
| |
| ' AIDS', |
| |
| 'Offline', |
| |
| ' Arcade', |
| |
| ' Unix', 'Unix', |
| |
| 'Network', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'internet-family', 1 |
|
|
| def prog_L06F21956(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.272). |
| Fires at 111,430 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "to" """ |
| |
| fire = { |
| |
| ' to', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'to-family', 1 |
|
|
| def prog_L06F21957(input_tokens): |
| """44 token families, 54 surface forms (train F1 of the uncapped optimum: 0.631). |
| Fires at 8,066 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to torrenting and file-sharing""" |
| |
| fire = { |
| |
| ' torrent', 'Torrent', ' Torrent', |
| |
| 'rent', ' Rent', |
| |
| ' fin', |
| |
| 'tor', |
| |
| ' pir', ' Pir', |
| |
| ' tracker', ' Tracker', 'Tracker', |
| |
| ' Allan', |
| |
| ' extradition', |
| |
| ' downloads', |
| |
| ' rainfall', |
| |
| '[/', ' [/', |
| |
| ' clamp', |
| |
| ' Dropbox', |
| |
| ' fugitive', |
| |
| ' vinyl', |
| |
| ' Album', |
| |
| 'pload', |
| |
| 'abytes', |
| |
| ' takedown', |
| |
| ' extrad', |
| |
| 'rar', |
| |
| ' DMCA', |
| |
| ' François', |
| |
| ' Boyle', |
| |
| 'Mbps', |
| |
| ' disconnect', |
| |
| ' Hud', |
| |
| ' Uri', ' URI', |
| |
| ' diff', |
| |
| ' regist', |
| |
| 'asar', |
| |
| ' fug', |
| |
| ' antiv', |
| |
| 'ansom', |
| |
| ' ether', |
| |
| 'orr', |
| |
| ' Aval', |
| |
| ' deprive', |
| |
| ' anonym', |
| |
| ' MPEG', 'mpeg', |
| |
| ' depri', |
| |
| ' thumbnail', |
| |
| ' attachments', 'Attach', |
| |
| ' radiator', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'torrent-family', 1 |
|
|
| def prog_L06F21958(input_tokens): |
| """8 token families, 9 surface forms (train F1 of the uncapped optimum: 0.193). |
| Fires at 153,782 of 36,786,566 positions. |
| Fires at 0.42% 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: names of individuals involved in controversy or conflict""" |
| |
| fire = { |
| |
| ' Trump', |
| |
| ' Mr', |
| |
| ' Clinton', |
| |
| ' Obama', |
| |
| 'j', |
| |
| ' Sanders', |
| |
| ' Paul', |
| |
| ' Jones', 'Jones', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'trump-family', 1 |
|
|
| def prog_L06F21959(input_tokens): |
| """12 token families, 22 surface forms (train F1 of the uncapped optimum: 0.165). |
| Fires at 341,408 of 36,786,566 positions. |
| Fires at 0.93% 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 legal accusations and intentionality""" |
| |
| fire = { |
| |
| '-', |
| |
| ' �', |
| |
| 't', 'ét', |
| |
| 're', |
| |
| ' very', 'very', ' VERY', |
| |
| 'and', |
| |
| ' far', 'far', |
| |
| ' edit', ' edited', |
| |
| ' un', 'un', |
| |
| 'ly', |
| |
| ' quite', 'quite', 'Quite', ' Quite', |
| |
| 'c', ' c', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '--family', 1 |
|
|
| def prog_L06F21960(input_tokens): |
| """4 token families, 8 surface forms (train F1 of the uncapped optimum: 0.268). |
| Fires at 12,856 of 36,786,566 positions.""" |
| |
| """prev SAE description: the phrase "something else." """ |
| |
| fire = { |
| |
| ' else', ' Else', 'else', 'Else', |
| |
| ' elsewhere', |
| |
| ' otherwise', |
| |
| 'other', 'OTHER', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'else-family', 1 |
|
|
| def prog_L06F21961(input_tokens): |
| """5 token families, 5 surface forms (train F1 of the uncapped optimum: 0.044). |
| Fires at 24,887 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to sentencing and prison terms""" |
| |
| fire = { |
| |
| ' to', |
| |
| ' faces', |
| |
| ' worth', |
| |
| ' suspended', |
| |
| ' prison', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'to-family', 1 |
|
|
| def prog_L06F21962(input_tokens): |
| """36 token families, 45 surface forms (train F1 of the uncapped optimum: 0.778). |
| Fires at 6,547 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to Mars and related terms""" |
| |
| fire = { |
| |
| ' Mars', ' MAR', 'Mars', ' mars', |
| |
| ' lunar', ' Lunar', |
| |
| ' rover', |
| |
| ' planetary', ' Planetary', |
| |
| ' planets', |
| |
| ' Martian', |
| |
| ' Mast', ' mast', 'mast', |
| |
| ' Rails', |
| |
| ' Venus', |
| |
| ' Curiosity', |
| |
| ' Antarctica', |
| |
| ' microbial', ' microbes', |
| |
| ' ISS', |
| |
| ' surpass', |
| |
| ' extrater', |
| |
| ' deport', |
| |
| 'oplan', |
| |
| ' Terra', |
| |
| ' sclerosis', |
| |
| ' macOS', |
| |
| ' STEM', 'STEM', |
| |
| ' prost', |
| |
| ' Deus', |
| |
| 'MRI', |
| |
| ' microbiome', |
| |
| 'arations', |
| |
| ' ALS', |
| |
| ' Ceres', |
| |
| 'arus', |
| |
| ' Mare', |
| |
| ' remed', |
| |
| 'gress', |
| |
| ' Petra', |
| |
| ' Kush', |
| |
| ' metast', |
| |
| ' Ares', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'mar-family', 1 |
|
|
| def prog_L06F21963(input_tokens): |
| """49 token families, 55 surface forms (train F1 of the uncapped optimum: 0.681). |
| Fires at 9,915 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "na" and its variations, suggesting a focus on phrases that describe a state or condition""" |
| |
| fire = { |
| |
| ' na', 'Na', |
| |
| ' bu', |
| |
| ' para', ' paras', |
| |
| ' ya', ' Ya', |
| |
| 'opa', |
| |
| ' belly', |
| |
| 'olla', |
| |
| 'aney', |
| |
| ' mas', |
| |
| ' Naw', |
| |
| ' mi', |
| |
| ' Ana', ' ana', |
| |
| ' ra', |
| |
| ' Brennan', |
| |
| 'sta', ' STA', |
| |
| ' pa', |
| |
| 'oca', |
| |
| 'osta', |
| |
| 'enei', |
| |
| ' Navajo', |
| |
| ' pil', |
| |
| ' Opera', |
| |
| ' Nou', |
| |
| ' underpin', |
| |
| ' Loch', |
| |
| ' validity', |
| |
| ' ye', |
| |
| ' Noon', |
| |
| ' dat', |
| |
| ' ng', |
| |
| ' Sana', |
| |
| ' Donn', |
| |
| ' Nah', 'nah', |
| |
| ' PowerShell', |
| |
| ' si', |
| |
| 'Aw', |
| |
| ' yo', |
| |
| ' ta', |
| |
| ' Nass', |
| |
| 'bek', |
| |
| ' clo', |
| |
| ' decon', |
| |
| ' pier', |
| |
| 'sein', |
| |
| ' tha', |
| |
| ' Sop', |
| |
| 'ajo', |
| |
| ' Nay', |
| |
| ' ni', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'na-family', 1 |
|
|
| def prog_L06F21964(input_tokens): |
| """15 token families, 20 surface forms (train F1 of the uncapped optimum: 0.250). |
| Fires at 107,577 of 36,786,566 positions.""" |
| |
| """prev SAE description: proper nouns, specifically names of people""" |
| |
| fire = { |
| |
| ' F', |
| |
| ' G', |
| |
| ' Smith', |
| |
| ' O', |
| |
| ' Brown', |
| |
| ' E', |
| |
| ' Jones', |
| |
| ' Mc', |
| |
| ' Van', ' van', |
| |
| ' De', |
| |
| ' Miller', ' Mills', ' Mill', |
| |
| ' Williams', |
| |
| ' Sch', ' SCH', ' Scher', |
| |
| ' McC', |
| |
| ' Johnson', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'f-family', 1 |
|
|
| def prog_L06F21965(input_tokens): |
| """45 token families, 59 surface forms (train F1 of the uncapped optimum: 0.643). |
| Fires at 11,103 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to dietary supplements""" |
| |
| fire = { |
| |
| ' supplement', ' supplements', 'Supplement', ' Supplement', |
| |
| ' supplies', |
| |
| ' bolster', |
| |
| ' expedition', ' Expedition', |
| |
| ' capsule', |
| |
| ' augmented', |
| |
| ' doping', |
| |
| ' dossier', |
| |
| ' Norton', |
| |
| ' pills', |
| |
| ' reprinted', ' reprint', |
| |
| ' Nunes', |
| |
| ' supplemental', ' Supplemental', ' supplemented', |
| |
| 'ADA', |
| |
| ' booster', ' Booster', ' boosters', |
| |
| ' steroids', ' steroid', |
| |
| 'lements', 'lement', |
| |
| ' Patreon', |
| |
| ' Berman', |
| |
| ' Omega', |
| |
| ' Swanson', |
| |
| ' tabletop', |
| |
| ' Suzanne', |
| |
| ' TC', |
| |
| ' ETF', 'ETF', |
| |
| ' supers', ' SUP', |
| |
| ' Nutrition', 'nutrition', |
| |
| ' bolstered', |
| |
| ' deficiency', |
| |
| ' Phelps', |
| |
| ' Olympia', |
| |
| ' adjunct', |
| |
| ' deficiencies', |
| |
| ' zinc', |
| |
| ' Defenders', |
| |
| ' mushroom', |
| |
| 'rint', |
| |
| 'endium', |
| |
| ' augment', |
| |
| ' booklet', |
| |
| ' powdered', |
| |
| 'reon', |
| |
| 'iciency', |
| |
| ' stocking', |
| |
| ' Garmin', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'supple-family', 1 |
|
|
| def prog_L06F21966(input_tokens): |
| """11 token families, 42 surface forms (train F1 of the uncapped optimum: 0.153). |
| Fires at 452,785 of 36,786,566 positions. |
| Fires at 1.23% 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: legal terminology and concepts""" |
| |
| fire = { |
| |
| ' to', |
| |
| ' work', ' working', ' works', ' worked', 'work', 'working', 'works', ' workings', ' WORK', 'workers', 'worked', |
| |
| ' it', |
| |
| ' years', 'year', 'years', ' YEAR', |
| |
| ' you', |
| |
| ' time', |
| |
| ' project', ' projects', ' Project', ' projected', 'project', ' Projects', |
| |
| ' up', |
| |
| ' people', |
| |
| ' designed', ' design', ' designing', ' Design', ' designs', ' designer', 'designed', ' Designer', ' Designed', |
| |
| ' charges', ' charged', 'charged', ' charger', 'charges', 'Charges', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'to-family', 1 |
|
|
| def prog_L06F21967(input_tokens): |
| """7 token families, 17 surface forms (train F1 of the uncapped optimum: 0.070). |
| Fires at 117,480 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to government and political themes""" |
| |
| fire = { |
| |
| ' has', |
| |
| ' said', |
| |
| ' announced', ' announces', |
| |
| ' officials', ' officially', |
| |
| ' government', ' governments', 'Government', ' Governments', ' governs', |
| |
| ' say', ' Says', ' Say', |
| |
| ' official', ' offices', 'official', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'has-family', 1 |
|
|
| def prog_L06F21968(input_tokens): |
| """45 token families, 51 surface forms (train F1 of the uncapped optimum: 0.618). |
| Fires at 7,806 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to advisory roles and councils""" |
| |
| fire = { |
| |
| ' advisory', ' Advisory', |
| |
| ' consulting', ' Consulting', |
| |
| ' advisors', ' Advisor', |
| |
| ' Austrian', |
| |
| 'ogenic', |
| |
| ' Important', 'Important', |
| |
| ' Allison', |
| |
| ' Monica', |
| |
| 'governmental', |
| |
| ' Advis', ' advis', |
| |
| ' therapeutic', |
| |
| 'ayan', |
| |
| 'aran', |
| |
| ' accessory', |
| |
| ' honorary', |
| |
| ' advising', ' advises', |
| |
| 'oro', |
| |
| ' Solutions', |
| |
| 'otive', |
| |
| ' é', |
| |
| 'ilateral', |
| |
| ' chaired', |
| |
| ' bulletin', |
| |
| 'erey', |
| |
| ' eclectic', |
| |
| ' informative', |
| |
| 'anian', |
| |
| ' Directors', |
| |
| ' authoritative', |
| |
| ' consultancy', |
| |
| ' informational', |
| |
| ' Expert', |
| |
| 'anical', |
| |
| ' sandy', |
| |
| 'visory', |
| |
| ' ay', |
| |
| ' Editorial', |
| |
| 'binding', |
| |
| 'agency', |
| |
| ' ministerial', |
| |
| ' Denise', |
| |
| ' brokerage', |
| |
| ' discretionary', |
| |
| ' observational', |
| |
| ' instructional', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'advisory-family', 1 |
|
|
| def prog_L06F21969(input_tokens): |
| """45 token families, 63 surface forms (train F1 of the uncapped optimum: 0.644). |
| Fires at 33,313 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to imaging and measurement technologies""" |
| |
| fire = { |
| |
| 'ographer', 'ograph', 'ographic', 'ographers', 'ographical', 'ographs', 'ographed', 'ographies', |
| |
| 'ography', |
| |
| 'ometer', 'ometers', |
| |
| 'ocrat', 'ocrats', 'ocratic', 'ocrates', |
| |
| 'ology', |
| |
| 'osphere', |
| |
| 'ivity', |
| |
| 'oğan', 'oan', |
| |
| 'ocese', |
| |
| 'olithic', 'olith', |
| |
| 'ophobia', |
| |
| 'ogram', 'ograms', |
| |
| 'imeter', 'imeters', 'imet', |
| |
| 'ophobic', 'ophob', |
| |
| 'ozy', |
| |
| ' Zuckerberg', |
| |
| 'oha', |
| |
| 'iveness', |
| |
| 'owitz', |
| |
| 'ogenic', 'ogenous', |
| |
| 'otechnology', |
| |
| 'atra', |
| |
| 'athon', |
| |
| 'ocracy', |
| |
| 'rome', |
| |
| ' exceptionally', |
| |
| 'utra', |
| |
| 'osc', |
| |
| 'awatts', |
| |
| 'plane', |
| |
| 'posium', |
| |
| 'abytes', |
| |
| 'ORPG', |
| |
| 'ophile', |
| |
| 'rones', |
| |
| 'opia', |
| |
| 'opoly', |
| |
| 'oise', |
| |
| 'atham', |
| |
| 'obia', |
| |
| 'eering', |
| |
| 'orama', |
| |
| ' screenplay', |
| |
| 'partisan', |
| |
| 'opathy', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ograph-family', 1 |
|
|
| def prog_L06F21970(input_tokens): |
| """25 token families, 31 surface forms (train F1 of the uncapped optimum: 0.740). |
| Fires at 6,139 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to virtual reality environments and concepts""" |
| |
| fire = { |
| |
| ' virtual', ' Virtual', 'Virtual', 'virtual', |
| |
| ' Ultimate', 'Ultimate', |
| |
| ' sy', |
| |
| ' videog', |
| |
| ' augmented', |
| |
| ' immersive', |
| |
| ' holog', |
| |
| ' sam', |
| |
| ' vouchers', ' voucher', |
| |
| ' kios', |
| |
| ' Vag', 'vag', |
| |
| 'hern', |
| |
| ' vending', |
| |
| ' avatar', |
| |
| ' simulated', |
| |
| ' pav', |
| |
| 'omorphic', |
| |
| 'raphic', |
| |
| ' Lunar', |
| |
| 'atars', |
| |
| 'bek', |
| |
| 'virt', |
| |
| ' phantom', |
| |
| 'resent', |
| |
| 'vik', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'virtu-family', 1 |
|
|
| def prog_L06F21971(input_tokens): |
| """15 token families, 34 surface forms (train F1 of the uncapped optimum: 0.226). |
| Fires at 1,154,246 of 36,786,566 positions. |
| Fires at 3.14% 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: newsworthy political events and discussions""" |
| |
| fire = { |
| |
| ' a', 'A', ' A', |
| |
| ' has', |
| |
| ' after', |
| |
| ' have', |
| |
| ' been', |
| |
| ' week', ' weeks', 'week', ' Weeks', |
| |
| ' year', ' years', |
| |
| ' two', 'Two', ' Two', |
| |
| ' new', ' newly', ' Newly', |
| |
| ' month', ' months', 'month', ' Months', |
| |
| ' over', |
| |
| ' days', ' day', 'day', 'Days', |
| |
| ' presidential', ' President', |
| |
| 'WASHINGTON', ' Washington', 'Washington', |
| |
| ' Thursday', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'a-family', 1 |
|
|
| def prog_L06F21972(input_tokens): |
| """16 token families, 23 surface forms (train F1 of the uncapped optimum: 0.769). |
| Fires at 16,791 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the word "questions." """ |
| |
| fire = { |
| |
| ' questions', ' questioning', ' Questions', 'Questions', |
| |
| ' question', ' Question', 'Question', 'question', |
| |
| ' wondering', |
| |
| ' FAQ', 'FAQ', |
| |
| ' doubts', |
| |
| ' queries', |
| |
| ' quiz', |
| |
| ' inquiries', |
| |
| ' quizz', |
| |
| ' AMA', |
| |
| ' questionnaire', |
| |
| ' inqu', |
| |
| ' trivia', |
| |
| ' interviewer', |
| |
| ' Grassley', |
| |
| ' misconceptions', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'question-family', 1 |
|
|
| def prog_L06F21973(input_tokens): |
| """31 token families, 36 surface forms (train F1 of the uncapped optimum: 0.830). |
| Fires at 5,811 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the term "crony" in various contexts related to economic and political critique""" |
| |
| fire = { |
| |
| ' cr', ' Cr', 'cr', 'Cr', |
| |
| ' Ec', ' ec', |
| |
| ' crust', |
| |
| ' Bryce', |
| |
| ' 2022', |
| |
| ' Cron', |
| |
| ' crane', |
| |
| ' crank', |
| |
| 'ocr', |
| |
| ' Rica', |
| |
| ' cringe', |
| |
| ' crater', |
| |
| ' bureaucr', |
| |
| ' pseud', |
| |
| ' circ', |
| |
| ' Geh', |
| |
| ' protr', |
| |
| ' overth', |
| |
| ' appar', |
| |
| ' desc', |
| |
| ' cradle', |
| |
| ' extr', |
| |
| ' crumbling', |
| |
| ' Pse', ' pse', |
| |
| ' crumble', |
| |
| 'ryce', |
| |
| ' occ', |
| |
| ' curl', |
| |
| ' eb', |
| |
| ' plaster', |
| |
| ' 236', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'cr-family', 1 |
|
|
| def prog_L06F21974(input_tokens): |
| """46 token families, 55 surface forms (train F1 of the uncapped optimum: 0.721). |
| Fires at 11,192 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the term "sid" and its variations in context""" |
| |
| fire = { |
| |
| ' Sid', ' sid', 'sided', 'sid', |
| |
| ' Tayyip', |
| |
| ' Trad', ' trad', |
| |
| ' tid', ' Tid', |
| |
| ' Ped', |
| |
| ' Tyson', |
| |
| ' Theft', |
| |
| 'vind', |
| |
| ' petty', ' Petty', |
| |
| ' Sanford', |
| |
| 'Background', |
| |
| 'inda', |
| |
| ' Gavin', |
| |
| ' Sidney', |
| |
| 'hardt', |
| |
| ' Brid', ' brid', |
| |
| 'idi', |
| |
| ' pedd', |
| |
| ' spac', ' Spac', |
| |
| ' tang', |
| |
| 'ando', |
| |
| 'stad', |
| |
| ' Din', |
| |
| ' obsc', |
| |
| ' Sett', 'sett', |
| |
| ' redist', |
| |
| ' rend', |
| |
| ' dialect', |
| |
| ' handic', |
| |
| ' Sands', |
| |
| 'buck', |
| |
| 'idad', |
| |
| ' Sidd', |
| |
| ' Butt', |
| |
| '380', |
| |
| ' sandbox', |
| |
| 'ippy', |
| |
| ' Desmond', |
| |
| ' Gri', |
| |
| ' polit', |
| |
| ' Tav', |
| |
| ' Trop', |
| |
| ' Sadd', |
| |
| ' gast', |
| |
| ' Aster', |
| |
| ' Brune', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'sid-family', 1 |
|
|
| def prog_L06F21975(input_tokens): |
| """8 token families, 14 surface forms (train F1 of the uncapped optimum: 0.069). |
| Fires at 83,399 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to the action of launching initiatives or campaigns""" |
| |
| fire = { |
| |
| ' a', |
| |
| ' an', |
| |
| ' its', |
| |
| ' new', |
| |
| ' launch', ' launched', ' launching', ' launches', 'launch', |
| |
| ' campaign', ' campaigns', |
| |
| ' effort', ' efforts', |
| |
| ' major', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'a-family', 1 |
|
|
| def prog_L06F21976(input_tokens): |
| """13 token families, 14 surface forms (train F1 of the uncapped optimum: 0.172). |
| Fires at 122,916 of 36,786,566 positions.""" |
| |
| """prev SAE description: proper nouns, particularly names and surnames""" |
| |
| fire = { |
| |
| 'man', 'mans', |
| |
| 'i', |
| |
| 'o', |
| |
| 'on', |
| |
| 'er', |
| |
| 'ie', |
| |
| 'en', |
| |
| 'ley', |
| |
| 'an', |
| |
| 'is', |
| |
| 'z', |
| |
| 'ner', |
| |
| 'son', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'man-family', 1 |
|
|
| def prog_L06F21977(input_tokens): |
| """7 token families, 12 surface forms (train F1 of the uncapped optimum: 0.144). |
| Fires at 31,027 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases that indicate struggling with various personal or mental health challenges""" |
| |
| fire = { |
| |
| ' with', |
| |
| ' facing', ' faced', |
| |
| ' battling', ' battled', |
| |
| ' experiencing', |
| |
| ' overcome', |
| |
| ' endured', ' enduring', |
| |
| ' confront', ' confronting', ' confronts', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'with-family', 1 |
|
|
| def prog_L06F21978(input_tokens): |
| """34 token families, 40 surface forms (train F1 of the uncapped optimum: 0.561). |
| Fires at 7,491 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to medications and their prescriptions""" |
| |
| fire = { |
| |
| ' drugs', |
| |
| ' medicine', |
| |
| ' pills', ' pill', |
| |
| ' pharmaceutical', ' Pharmaceutical', |
| |
| ' medication', |
| |
| ' treatments', |
| |
| ' medications', |
| |
| ' psychiatrist', ' psychiatrists', |
| |
| ' therapy', |
| |
| ' medicines', |
| |
| ' prescribed', |
| |
| ' vaccine', |
| |
| 'killers', 'killer', |
| |
| ' antidepressants', ' antidepressant', |
| |
| 'imen', 'imens', |
| |
| 'ratom', |
| |
| ' counseling', |
| |
| 'heter', |
| |
| ' regimen', |
| |
| ' narcotics', |
| |
| ' morphine', |
| |
| ' neurotrans', |
| |
| ' Lydia', |
| |
| ' aspirin', |
| |
| ' bipolar', |
| |
| ' interventions', |
| |
| ' psychiatry', |
| |
| ' hypertension', |
| |
| 'idine', |
| |
| ' prescribe', |
| |
| 'agra', |
| |
| ' workaround', |
| |
| 'ophen', |
| |
| ' dosage', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'drug-family', 1 |
|
|
| def prog_L06F21979(input_tokens): |
| """21 token families, 35 surface forms (train F1 of the uncapped optimum: 0.222). |
| Fires at 43,972 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to career-high statistics in sports""" |
| |
| fire = { |
| |
| ' high', ' highest', 'high', ' highs', 'highest', ' HIGH', ' Highest', |
| |
| ' record', 'record', ' Record', |
| |
| ' champion', ' Champion', ' champions', |
| |
| ' double', 'double', |
| |
| ' largest', |
| |
| ' total', |
| |
| ' straight', 'straight', |
| |
| ' gold', |
| |
| 'winning', 'winner', |
| |
| ' consecutive', |
| |
| ' milestone', |
| |
| ' longest', |
| |
| ' multi', |
| |
| ' crimes', |
| |
| 'breaking', |
| |
| ' triple', ' Triple', |
| |
| ' medal', |
| |
| ' whopping', |
| |
| ' streak', |
| |
| 'best', |
| |
| ' fastest', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'high-family', 1 |
|
|
| def prog_L06F21980(input_tokens): |
| """20 token families, 37 surface forms (train F1 of the uncapped optimum: 0.315). |
| Fires at 52,689 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to embodiment, manifestation, and incarnation, often in political or philosophical contexts""" |
| |
| fire = { |
| |
| ' version', 'version', |
| |
| ' example', |
| |
| ' product', 'product', 'products', |
| |
| ' victim', |
| |
| ' representative', ' representatives', |
| |
| ' adaptation', |
| |
| ' shadow', ' shadows', 'shadow', |
| |
| ' versions', 'versions', |
| |
| ' reflection', |
| |
| ' symbol', ' symbols', |
| |
| ' examples', ' Examples', 'Examples', |
| |
| ' expression', 'expression', |
| |
| ' extension', |
| |
| ' height', |
| |
| ' iteration', |
| |
| ' blow', |
| |
| ' echo', ' echoes', 'echo', ' echoing', |
| |
| ' essence', 'esses', ' Essence', 'essment', |
| |
| ' definition', |
| |
| ' portrait', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ver-family', 1 |
|
|
| def prog_L06F21981(input_tokens): |
| """19 token families, 34 surface forms (train F1 of the uncapped optimum: 0.740). |
| Fires at 11,393 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to prisons and incarceration""" |
| |
| fire = { |
| |
| ' prison', ' prisons', ' Prison', ' prisoners', ' prisoner', 'prison', |
| |
| ' jail', ' Jail', ' jails', |
| |
| ' detention', ' Detention', |
| |
| ' imprisoned', ' imprisonment', ' imprison', |
| |
| ' Corrections', ' Correctional', ' correctional', ' corrections', |
| |
| ' Guantanamo', |
| |
| 'Pr', |
| |
| ' incarceration', |
| |
| ' penal', ' Penal', |
| |
| ' confinement', |
| |
| ' Arkham', |
| |
| 'ISON', 'isons', |
| |
| ' incarcerated', |
| |
| ' Auschwitz', |
| |
| 'ihar', |
| |
| ' Guant', |
| |
| ' PNG', |
| |
| ' Hogwarts', |
| |
| ' Labyrinth', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'prison-family', 1 |
|
|
| def prog_L06F21982(input_tokens): |
| """31 token families, 39 surface forms (train F1 of the uncapped optimum: 0.740). |
| Fires at 11,387 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "ag" in various contexts""" |
| |
| fire = { |
| |
| 'ig', 'IG', ' ig', |
| |
| 'rig', ' Rig', |
| |
| ' lig', 'lig', |
| |
| 'iger', |
| |
| 'oug', |
| |
| ' Miguel', |
| |
| 'igs', |
| |
| ' Gig', |
| |
| 'zig', ' Zig', |
| |
| ' wig', 'wig', |
| |
| 'iga', |
| |
| 'igo', |
| |
| ' Pig', |
| |
| ' Mig', ' mig', |
| |
| 'lich', |
| |
| 'igel', |
| |
| 'Dou', |
| |
| 'dig', |
| |
| ' Berk', |
| |
| 'igi', |
| |
| '__', |
| |
| 'igger', |
| |
| ' Nig', ' nig', |
| |
| 'igne', |
| |
| ' eyel', |
| |
| 'aber', |
| |
| 'igil', |
| |
| ' Fairy', |
| |
| '________', |
| |
| 'ー', |
| |
| 'ー�', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ig-family', 1 |
|
|
| def prog_L06F21983(input_tokens): |
| """34 token families, 35 surface forms (train F1 of the uncapped optimum: 0.451). |
| Fires at 2,931 of 36,786,566 positions.""" |
| |
| """prev SAE description: special characters or non-standard symbols in the text""" |
| |
| fire = { |
| |
| ' semif', |
| |
| 'iny', |
| |
| ' Lowe', |
| |
| 'м', |
| |
| ' Dup', |
| |
| ' finalists', |
| |
| '————', |
| |
| ' Kle', |
| |
| ' Supplemental', |
| |
| 'Cut', |
| |
| 'nette', |
| |
| 'τ', |
| |
| ' Sug', |
| |
| ' incom', |
| |
| ' Sponge', |
| |
| ' shrunk', |
| |
| ' Slightly', |
| |
| ' UNC', |
| |
| '————————', |
| |
| 'て', |
| |
| 'uper', |
| |
| ' ∼', |
| |
| 'ト', |
| |
| ' ↑', '↑', |
| |
| 'roxy', |
| |
| ' >=', |
| |
| 'し', |
| |
| 'ILLE', |
| |
| 'プ', |
| |
| ' centimeters', |
| |
| 'Bonus', |
| |
| 'מ', |
| |
| '------------', |
| |
| ' ++', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'semif-family', 1 |
|
|
| def prog_L06F21984(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.549). |
| Fires at 28,116 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "one" with varying contexts indicating unity or singularity""" |
| |
| fire = { |
| |
| ' one', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'one-family', 1 |
|
|
| def prog_L06F21985(input_tokens): |
| """10 token families, 12 surface forms (train F1 of the uncapped optimum: 0.781). |
| Fires at 8,287 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "completely" and its variants in various contexts""" |
| |
| fire = { |
| |
| ' completely', 'completely', |
| |
| ' entirely', |
| |
| ' totally', ' Totally', |
| |
| ' thoroughly', |
| |
| ' fundamentally', |
| |
| ' utterly', |
| |
| ' drastically', |
| |
| ' radically', |
| |
| 'otally', |
| |
| 'handedly', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'complete-family', 1 |
|
|
| def prog_L06F21986(input_tokens): |
| """34 token families, 37 surface forms (train F1 of the uncapped optimum: 0.553). |
| Fires at 6,723 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to touchdowns in sports contexts""" |
| |
| fire = { |
| |
| ' touchdown', ' touchdowns', |
| |
| ' championship', ' championships', |
| |
| ' delicious', |
| |
| ' Test', |
| |
| ' tasty', |
| |
| ' homer', ' homers', |
| |
| ' TD', |
| |
| ' interceptions', |
| |
| ' interception', |
| |
| ' Heisman', |
| |
| ' verb', |
| |
| ' THC', |
| |
| ' receptions', |
| |
| ' catchy', |
| |
| ' rebounds', |
| |
| 'iership', |
| |
| ' TDs', |
| |
| ' noun', |
| |
| ' ethanol', |
| |
| ' strikeouts', |
| |
| ' orgasm', |
| |
| 'Mbps', |
| |
| ' flavorful', |
| |
| ' fumble', |
| |
| ' dunk', |
| |
| ' melody', |
| |
| ' sunrise', |
| |
| ' ignition', |
| |
| ' timeout', |
| |
| 'pointers', |
| |
| ' melodies', |
| |
| ' climax', |
| |
| ' righteous', |
| |
| ' ABV', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'touchdown-family', 1 |
|
|
| def prog_L06F21987(input_tokens): |
| """16 token families, 16 surface forms (train F1 of the uncapped optimum: 0.191). |
| Fires at 24,421 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of specific names, particularly those of political figures""" |
| |
| fire = { |
| |
| ' Donald', |
| |
| ' John', |
| |
| ' |', |
| |
| ' (@', |
| |
| ' invalid', |
| |
| ' Whats', |
| |
| 'House', |
| |
| ' speaks', |
| |
| ' urges', |
| |
| ' Diane', |
| |
| ' Rodham', |
| |
| 'Sanders', |
| |
| ' Bernard', |
| |
| ' attends', |
| |
| 'Bernie', |
| |
| ' ·', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'donald-family', 1 |
|
|
| def prog_L06F21988(input_tokens): |
| """47 token families, 56 surface forms (train F1 of the uncapped optimum: 0.590). |
| Fires at 20,675 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to destruction or damage""" |
| |
| fire = { |
| |
| ' wreck', 'wreck', ' wrecked', |
| |
| ' embark', ' embarked', |
| |
| ' parked', |
| |
| ' unleashed', ' unleash', |
| |
| ' stranded', |
| |
| ' ruins', ' ruin', |
| |
| ' Shares', |
| |
| ' inflicted', ' inflict', ' inflicting', |
| |
| ' footprint', ' footprints', |
| |
| ' spree', |
| |
| ' beast', |
| |
| ' rampage', |
| |
| ' havoc', |
| |
| ' wre', |
| |
| ' churn', |
| |
| ' slew', |
| |
| 'reck', ' Reck', |
| |
| ' devastation', |
| |
| ' battered', |
| |
| ' wreckage', |
| |
| ' spew', |
| |
| ' Zack', |
| |
| 'cks', |
| |
| ' devastated', |
| |
| ' onslaught', |
| |
| ' Tucker', |
| |
| ' wiped', |
| |
| ' stre', |
| |
| ' mayhem', |
| |
| ' rav', |
| |
| ' crater', |
| |
| ' littered', |
| |
| ' monstrous', |
| |
| ' racked', |
| |
| ' crippled', |
| |
| ' carnage', |
| |
| ' stomp', |
| |
| ' swath', |
| |
| ' ravaged', |
| |
| ' Bedford', |
| |
| ' scourge', |
| |
| ' perpetrated', |
| |
| ' riddled', |
| |
| ' ashore', |
| |
| 'estead', |
| |
| ' dere', |
| |
| ' poop', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'wreck-family', 1 |
|
|
| def prog_L06F21989(input_tokens): |
| """10 token families, 15 surface forms (train F1 of the uncapped optimum: 0.057). |
| Fires at 44,581 of 36,786,566 positions.""" |
| |
| """prev SAE description: the end of sentences or paragraphs""" |
| |
| fire = { |
| |
| ' new', ' NEW', |
| |
| ' man', ' MAN', |
| |
| ' group', ' GROUP', ' grouping', |
| |
| ' former', |
| |
| ' woman', |
| |
| ' study', ' Study', |
| |
| ' friend', |
| |
| ' film', |
| |
| ' couple', |
| |
| ' guest', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'new-family', 1 |
|
|
| def prog_L06F21990(input_tokens): |
| """26 token families, 52 surface forms (train F1 of the uncapped optimum: 0.239). |
| Fires at 40,088 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to competitive rankings or awards, especially in the context of individuals or entities being recognized""" |
| |
| fire = { |
| |
| 'up', |
| |
| ' winner', ' winners', ' Winner', 'winner', ' Winners', 'Winner', |
| |
| ' champion', ' Champion', ' champions', |
| |
| 'elect', |
| |
| 'runner', ' runner', 'runners', ' runners', 'Runner', |
| |
| ' entry', ' Entry', |
| |
| ' Year', |
| |
| ' overall', |
| |
| ' selection', ' Selection', 'selection', |
| |
| ' mate', 'mate', |
| |
| 'Star', 'Stars', 'stars', |
| |
| ' medal', |
| |
| ' Player', |
| |
| 'lining', 'liner', 'liners', |
| |
| ' seed', 'seed', ' seeded', |
| |
| ' row', 'rower', |
| |
| 'ring', |
| |
| ' prize', |
| |
| ' Choice', 'choice', |
| |
| 'card', 'cards', |
| |
| 'uts', |
| |
| 'iser', |
| |
| ' contenders', ' contender', |
| |
| 'hard', |
| |
| ' entries', |
| |
| ' holder', 'holder', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'up-family', 1 |
|
|
| def prog_L06F21991(input_tokens): |
| """22 token families, 30 surface forms (train F1 of the uncapped optimum: 0.695). |
| Fires at 7,353 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to "feet" in various contexts""" |
| |
| fire = { |
| |
| ' feet', 'feet', ' Feet', |
| |
| ' foot', 'foot', ' footing', 'footed', 'Foot', |
| |
| ' boots', ' Boots', |
| |
| ' ankle', |
| |
| ' footprint', ' footprints', |
| |
| ' Miles', |
| |
| ' footsteps', |
| |
| ' stairs', |
| |
| ' socks', |
| |
| ' floors', |
| |
| ' toes', |
| |
| ' doorstep', |
| |
| ' foothold', |
| |
| ' stride', |
| |
| ' fingertips', |
| |
| 'fts', |
| |
| ' ankles', |
| |
| ' pH', |
| |
| ' paws', |
| |
| 'ranean', |
| |
| ' Misty', |
| |
| ' grounding', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'feet-family', 1 |
|
|
| def prog_L06F21992(input_tokens): |
| """50 token families, 52 surface forms (train F1 of the uncapped optimum: 0.668). |
| Fires at 14,683 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to international locations and legal contexts""" |
| |
| fire = { |
| |
| ' Switzerland', |
| |
| ' Geneva', |
| |
| ' Damascus', |
| |
| ' Vienna', |
| |
| ' Haven', 'haven', |
| |
| ' Hague', |
| |
| ' Durham', |
| |
| ' Lavrov', |
| |
| ' Sergei', |
| |
| 'Gene', |
| |
| ' Oslo', |
| |
| ' Godzilla', |
| |
| ' Wembley', |
| |
| ' Calais', |
| |
| ' Florence', |
| |
| ' Windsor', |
| |
| ' IL', |
| |
| ' Millennium', |
| |
| ' Heller', |
| |
| ' gases', |
| |
| ' Riyadh', |
| |
| 'eca', 'ECA', |
| |
| 'GEN', |
| |
| ' specifics', |
| |
| ' Cannes', |
| |
| ' Zurich', |
| |
| ' infrared', |
| |
| ' Copenhagen', |
| |
| ' Idlib', |
| |
| ' Laurie', |
| |
| ' Dayton', |
| |
| ' Goodell', |
| |
| 'essa', |
| |
| ' Hartford', |
| |
| ' Norris', |
| |
| ' Kyoto', |
| |
| ' Lisbon', |
| |
| ' Stras', |
| |
| ' Hyde', |
| |
| ' 1949', |
| |
| ' Louise', |
| |
| ' Lancet', |
| |
| 'EVA', |
| |
| ' Helsinki', |
| |
| 'insk', |
| |
| ' Cornwall', |
| |
| ' Hiroshima', |
| |
| ' Naomi', |
| |
| ' indifference', |
| |
| ' Dresden', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'switzerland-family', 1 |
|
|
| def prog_L06F21993(input_tokens): |
| """5 token families, 7 surface forms (train F1 of the uncapped optimum: 0.111). |
| Fires at 46,672 of 36,786,566 positions.""" |
| |
| """prev SAE description: pronouns, particularly "they" and "it," indicating a focus on subjects and their relationships in statements""" |
| |
| fire = { |
| |
| ' it', |
| |
| ' they', 'they', |
| |
| ' he', |
| |
| ' she', |
| |
| ' fear', ' feared', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'it-family', 1 |
|
|
| def prog_L06F21994(input_tokens): |
| """41 token families, 59 surface forms (train F1 of the uncapped optimum: 0.588). |
| Fires at 19,382 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to praise or acclaim""" |
| |
| fire = { |
| |
| ' hailed', ' hail', ' Hail', |
| |
| ' blasted', ' blasting', |
| |
| 'awaited', ' awaited', |
| |
| ' welcomed', |
| |
| ' touted', ' touting', |
| |
| ' praised', |
| |
| ' acclaimed', ' acclaim', |
| |
| ' proclaimed', ' proclaiming', ' proclaim', 'proclaimed', |
| |
| ' celebrated', |
| |
| 'ailing', 'ailed', |
| |
| ' greeted', |
| |
| ' billed', ' billing', |
| |
| ' herald', |
| |
| ' glorious', |
| |
| ' lauded', ' laud', |
| |
| ' slammed', |
| |
| ' crowned', |
| |
| 'Overview', |
| |
| ' roaring', ' roar', ' roared', |
| |
| ' shaping', |
| |
| 'Calling', ' Calling', |
| |
| ' Mighty', ' mighty', |
| |
| ' oft', |
| |
| ' denounced', ' denouncing', |
| |
| ' reeling', |
| |
| ' breakout', |
| |
| ' championed', |
| |
| ' accol', |
| |
| 'reetings', |
| |
| ' Celebr', ' celebr', |
| |
| ' cheered', |
| |
| 'ouncing', |
| |
| ' triumphant', |
| |
| ' lamented', |
| |
| ' renaissance', |
| |
| ' praises', |
| |
| ' cele', |
| |
| ' payday', |
| |
| ' Cohn', |
| |
| 'Synopsis', |
| |
| ' rejoice', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'hail-family', 1 |
|
|
| def prog_L06F21995(input_tokens): |
| """36 token families, 40 surface forms (train F1 of the uncapped optimum: 0.756). |
| Fires at 5,941 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to lists or items in a structured format""" |
| |
| fire = { |
| |
| 'li', 'LI', ' li', |
| |
| ' •', |
| |
| 'lli', |
| |
| 'oni', |
| |
| ' Eli', 'eli', |
| |
| ' Lem', |
| |
| ' Colts', |
| |
| ' Father', |
| |
| 'oli', |
| |
| ' Liu', |
| |
| ' wholly', |
| |
| ' Gan', |
| |
| ' Lionel', |
| |
| 'lie', |
| |
| ' Liga', 'liga', |
| |
| 'oshi', |
| |
| ' stre', |
| |
| ' Tolkien', |
| |
| 'ti', |
| |
| ' Ll', |
| |
| 'tti', |
| |
| ' DeL', |
| |
| ' Sweeney', |
| |
| ' Zhang', |
| |
| ' Levi', |
| |
| ' 330', |
| |
| 'ogh', |
| |
| ' Manit', |
| |
| ' Lei', |
| |
| ' cabbage', |
| |
| 'lv', |
| |
| ' Liver', |
| |
| ' vertically', |
| |
| ' Liang', |
| |
| ' Eliot', |
| |
| ' Fraz', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'li-family', 1 |
|
|
| def prog_L06F21996(input_tokens): |
| """18 token families, 42 surface forms (train F1 of the uncapped optimum: 0.158). |
| Fires at 191,560 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: reports that contain statistical data or findings on social issues""" |
| |
| fire = { |
| |
| ' that', |
| |
| ' report', ' Report', |
| |
| ' released', 'released', 'Released', |
| |
| ' found', |
| |
| ' shows', ' showed', ' showing', 'shows', |
| |
| ' study', ' Study', 'study', |
| |
| ' announced', |
| |
| ' says', |
| |
| ' published', |
| |
| ' Wednesday', |
| |
| ' warned', ' warning', ' warns', ' warn', |
| |
| ' research', ' researchers', 'Researchers', ' Researchers', 'Research', |
| |
| ' suggests', ' suggested', ' suggesting', |
| |
| ' revealed', ' reveals', ' revealing', |
| |
| ' issued', |
| |
| ' finds', ' findings', 'finding', |
| |
| ' concluded', ' concludes', ' concluding', |
| |
| ' annual', ' Annual', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'that-family', 1 |
|
|
| def prog_L06F21997(input_tokens): |
| """2 token families, 2 surface forms (train F1 of the uncapped optimum: 0.277). |
| Fires at 2,576 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the website "twitter.com" """ |
| |
| fire = { |
| |
| 'com', |
| |
| 'twitter', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'com-family', 1 |
|
|
| def prog_L06F21998(input_tokens): |
| """48 token families, 48 surface forms (train F1 of the uncapped optimum: 0.693). |
| Fires at 7,387 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the city of Mosul and its associated events or situations""" |
| |
| fire = { |
| |
| ' Mosul', |
| |
| ' Glasgow', |
| |
| ' Baghdad', |
| |
| ' airstrikes', |
| |
| ' Maps', |
| |
| ' Pluto', |
| |
| ' Ramadan', |
| |
| ' Nim', |
| |
| ' Katrina', |
| |
| ' Pesh', |
| |
| 'bil', |
| |
| ' rubble', |
| |
| ' Karachi', |
| |
| ' Yaz', |
| |
| ' airstrike', |
| |
| ' Abrams', |
| |
| ' Trafford', |
| |
| 'airo', |
| |
| ' Ryzen', |
| |
| ' UNESCO', |
| |
| ' Petraeus', |
| |
| ' rav', |
| |
| 'myra', |
| |
| 'aghd', |
| |
| ' Spears', |
| |
| ' Daesh', |
| |
| ' polio', |
| |
| ' retake', |
| |
| 'ujah', |
| |
| ' Manziel', |
| |
| ' Kun', |
| |
| 'agar', |
| |
| 'Mos', |
| |
| 'imation', |
| |
| 'ulia', |
| |
| ' Loch', |
| |
| 'iyah', |
| |
| ' mars', |
| |
| ' Geo', |
| |
| ' Kund', |
| |
| ' Hitman', |
| |
| ' Unicode', |
| |
| ' Shinji', |
| |
| 'ukong', |
| |
| ' 6000', |
| |
| ' Baghd', |
| |
| 'upiter', |
| |
| ' MUS', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'mosul-family', 1 |
|
|
| def prog_L06F21999(input_tokens): |
| """30 token families, 38 surface forms (train F1 of the uncapped optimum: 0.675). |
| Fires at 7,674 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to various types of cells in biological contexts, particularly immune and stem cells""" |
| |
| fire = { |
| |
| ' cells', ' cell', ' Cell', ' Cells', 'cells', |
| |
| ' organisms', ' organism', |
| |
| ' neurons', ' neuronal', ' neuron', |
| |
| ' embryos', ' embryo', |
| |
| ' galaxies', |
| |
| ' organs', |
| |
| ' cellular', |
| |
| ' lymph', |
| |
| ' coc', |
| |
| ' leukemia', |
| |
| ' nucle', |
| |
| ' regener', |
| |
| ' marrow', |
| |
| 'llular', |
| |
| ' RNA', |
| |
| 'obl', |
| |
| ' embryonic', |
| |
| ' buds', |
| |
| 'acteria', |
| |
| 'ubes', |
| |
| 'ocytes', |
| |
| 'ucle', |
| |
| ' differentiation', |
| |
| 'derived', |
| |
| ' clot', |
| |
| ' plur', |
| |
| ' regenerate', |
| |
| ' chromosomes', |
| |
| 'otent', |
| |
| 'helial', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'cell-family', 1 |
|
|
