| """GPT-2 SAE last-token programs, layer 6 -- features 08000 to 08099 (100 programs). |
| |
| One function per SAE feature. Grouped 100 to a file so the set stays browsable; |
| the programs themselves are unchanged. |
| """ |
|
|
| def prog_L06F08000(input_tokens): |
| """6 token families, 9 surface forms (train F1 of the uncapped optimum: 0.740). |
| Fires at 6,546 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "Follow" indicating social media references""" |
| |
| fire = { |
| |
| ' follow', ' Follow', 'Follow', 'follow', |
| |
| 'WATCH', |
| |
| 'Find', |
| |
| 'Contact', |
| |
| 'LLOW', |
| |
| 'OTO', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'follow-family', 1 |
|
|
| def prog_L06F08001(input_tokens): |
| """13 token families, 21 surface forms (train F1 of the uncapped optimum: 0.055). |
| Fires at 71,367 of 36,786,566 positions.""" |
| |
| """prev SAE description: expressions of strong sentiments or reactions""" |
| |
| fire = { |
| |
| ' a', |
| |
| ' an', |
| |
| ' some', |
| |
| ' them', 'them', |
| |
| ' given', 'given', |
| |
| ' away', 'aways', ' Away', |
| |
| ' free', ' Free', ' FREE', |
| |
| ' us', |
| |
| ' birth', ' Birth', |
| |
| ' following', |
| |
| ' details', ' detailed', |
| |
| ' information', |
| |
| ' little', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'a-family', 1 |
|
|
| def prog_L06F08002(input_tokens): |
| """37 token families, 48 surface forms (train F1 of the uncapped optimum: 0.764). |
| Fires at 5,367 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the company AMD and its products""" |
| |
| fire = { |
| |
| ' AMD', 'AMD', 'amd', ' amd', |
| |
| ' Oracle', 'Oracle', |
| |
| ' Motorola', |
| |
| ' ARM', 'ARM', |
| |
| ' ASUS', ' Asus', 'asus', |
| |
| ' Opera', |
| |
| ' Radeon', |
| |
| ' Qualcomm', |
| |
| 'VERTIS', |
| |
| ' MHz', 'MHz', |
| |
| 'nm', ' nm', |
| |
| ' Nvidia', 'NVIDIA', |
| |
| ' Ryzen', |
| |
| ' VMware', |
| |
| 'Nintendo', |
| |
| ' NOAA', |
| |
| ' MSI', |
| |
| ' MLG', |
| |
| ' Aust', |
| |
| ' motherboard', |
| |
| ' Acer', |
| |
| 'Intel', |
| |
| ' overcl', |
| |
| 'XT', |
| |
| 'BSD', |
| |
| ' integer', 'Integer', |
| |
| ' OEM', |
| |
| ' architectures', |
| |
| 'DEP', |
| |
| ' FreeBSD', |
| |
| ' Noct', |
| |
| 'vidia', |
| |
| 'adeon', |
| |
| ' CONTIN', |
| |
| ' Instruments', |
| |
| ' ATK', |
| |
| 'awei', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'amd-family', 1 |
|
|
| def prog_L06F08003(input_tokens): |
| """6 token families, 11 surface forms (train F1 of the uncapped optimum: 0.774). |
| Fires at 17,557 of 36,786,566 positions.""" |
| |
| """prev SAE description: conditional phrases that introduce contrasting or qualifying statements""" |
| |
| fire = { |
| |
| 'While', ' While', 'while', |
| |
| ' Although', 'Although', |
| |
| ' Though', |
| |
| ' whilst', ' Whilst', |
| |
| ' Whereas', 'Whereas', |
| |
| 'ilst', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'while-family', 1 |
|
|
| def prog_L06F08004(input_tokens): |
| """8 token families, 10 surface forms (train F1 of the uncapped optimum: 0.121). |
| Fires at 40,431 of 36,786,566 positions.""" |
| |
| """prev SAE description: names of prominent individuals, particularly those with first and last names starting with 'Robert' or 'C' and their associated titles""" |
| |
| fire = { |
| |
| ' K', |
| |
| ' van', ' Van', |
| |
| ' Mueller', |
| |
| ' Moore', |
| |
| ' De', |
| |
| ' Reagan', |
| |
| ' Fantasy', |
| |
| ' Mc', 'mc', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'k-family', 1 |
|
|
| def prog_L06F08005(input_tokens): |
| """35 token families, 43 surface forms (train F1 of the uncapped optimum: 0.720). |
| Fires at 7,225 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the name "Walker." """ |
| |
| fire = { |
| |
| ' Walker', ' Walk', 'Walker', 'walker', 'Walk', |
| |
| 'owski', |
| |
| 'lli', |
| |
| ' Elliott', |
| |
| ' Davidson', |
| |
| ' Guardiola', |
| |
| ' Cage', |
| |
| ' Kejriwal', |
| |
| 'idget', |
| |
| 'steen', |
| |
| 'andowski', |
| |
| 'ISON', |
| |
| ' Cannon', |
| |
| 'taker', |
| |
| ' Andersen', |
| |
| ' Archer', |
| |
| ' coil', ' coils', ' Coil', |
| |
| 'Keefe', |
| |
| ' Shelley', |
| |
| ' Rookie', |
| |
| ' Feldman', |
| |
| ' Patton', |
| |
| ' Williamson', |
| |
| 'urden', |
| |
| 'ALK', |
| |
| ' Walls', |
| |
| ' ALEC', |
| |
| 'ewski', |
| |
| 'Wilson', |
| |
| 'keleton', |
| |
| ' Worker', |
| |
| 'Cour', |
| |
| ' widget', 'widget', 'Widget', |
| |
| 'rique', |
| |
| ' Painter', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'walk-family', 1 |
|
|
| def prog_L06F08006(input_tokens): |
| """6 token families, 7 surface forms (train F1 of the uncapped optimum: 0.117). |
| Fires at 21,327 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to specific SCP (Secure, Contain, Protect) objects, particularly their designations and associated characteristics""" |
| |
| fire = { |
| |
| 'com', |
| |
| ' SCP', 'SCP', |
| |
| ' #', |
| |
| '_', |
| |
| ' Things', |
| |
| 'story', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'com-family', 1 |
|
|
| def prog_L06F08007(input_tokens): |
| """35 token families, 47 surface forms (train F1 of the uncapped optimum: 0.591). |
| Fires at 42,325 of 36,786,566 positions.""" |
| |
| """prev SAE description: words related to brand names and products""" |
| |
| fire = { |
| |
| 'aker', 'akers', |
| |
| 'pper', 'ppers', |
| |
| 'ipper', 'ippers', |
| |
| 'CLE', 'cle', 'cles', |
| |
| 'apper', 'appers', 'apps', |
| |
| 'andy', |
| |
| 'ancy', |
| |
| 'anta', |
| |
| 'oda', |
| |
| 'aver', 'avers', |
| |
| 'uba', |
| |
| 'acker', |
| |
| 'ikers', 'iker', |
| |
| 'assion', |
| |
| 'orter', 'orters', |
| |
| 'ava', |
| |
| 'oga', |
| |
| 'eds', |
| |
| 'inkle', |
| |
| 'ucker', |
| |
| 'aja', |
| |
| 'cs', |
| |
| 'uchin', |
| |
| 'udge', |
| |
| 'cker', |
| |
| 'uddle', |
| |
| 'ervatives', 'ervative', |
| |
| 'agle', |
| |
| 'amo', |
| |
| 'igator', 'igators', |
| |
| 'avan', |
| |
| 'istle', |
| |
| 'boat', |
| |
| 'utz', |
| |
| 'entry', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'aker-family', 1 |
|
|
| def prog_L06F08008(input_tokens): |
| """1 token families, 5 surface forms (train F1 of the uncapped optimum: 0.585). |
| Fires at 20,803 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "over" used in various contexts""" |
| |
| fire = { |
| |
| ' over', ' Over', 'over', ' OVER', 'OVER', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'over-family', 1 |
|
|
| def prog_L06F08009(input_tokens): |
| """17 token families, 19 surface forms (train F1 of the uncapped optimum: 0.297). |
| Fires at 7,424 of 36,786,566 positions.""" |
| |
| """prev SAE description: statistical references to averages and means""" |
| |
| fire = { |
| |
| ' average', 'average', |
| |
| ' imagine', |
| |
| ' default', |
| |
| 'surprisingly', |
| |
| ' masse', |
| |
| ' essence', |
| |
| ' contrary', |
| |
| ' Typically', 'Typically', |
| |
| ' capita', |
| |
| ' ideally', |
| |
| 'erous', |
| |
| ' aggregate', |
| |
| 'istically', |
| |
| ' statistically', |
| |
| 'prisingly', |
| |
| 'rox', |
| |
| ' averages', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'average-family', 1 |
|
|
| def prog_L06F08010(input_tokens): |
| """19 token families, 25 surface forms (train F1 of the uncapped optimum: 0.146). |
| Fires at 80,259 of 36,786,566 positions.""" |
| |
| """prev SAE description: directional language and phrases related to advantages and disadvantages""" |
| |
| fire = { |
| |
| ' end', |
| |
| ' level', ' levels', |
| |
| ' top', |
| |
| ' side', ' Side', |
| |
| ' death', |
| |
| ' point', |
| |
| ' ground', |
| |
| ' hospital', |
| |
| ' victory', |
| |
| ' date', 'date', |
| |
| ' floor', ' Floor', |
| |
| ' bottom', 'bottom', |
| |
| ' streets', |
| |
| ' safety', |
| |
| ' ballot', |
| |
| ' stage', |
| |
| ' edge', |
| |
| ' door', 'door', |
| |
| ' success', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'end-family', 1 |
|
|
| def prog_L06F08011(input_tokens): |
| """41 token families, 53 surface forms (train F1 of the uncapped optimum: 0.325). |
| Fires at 75,670 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms and phrases related to generalizations and overall assessments""" |
| |
| fire = { |
| |
| ' official', ' Official', |
| |
| ' federal', ' Federal', |
| |
| ' final', ' finer', |
| |
| ' national', |
| |
| ' general', |
| |
| ' main', |
| |
| ' personal', |
| |
| ' total', ' TOTAL', |
| |
| ' original', |
| |
| ' entire', |
| |
| ' global', |
| |
| ' actual', ' Actual', 'actual', |
| |
| ' overall', |
| |
| ' daily', |
| |
| ' basic', 'basic', ' basal', |
| |
| ' initial', |
| |
| ' vast', |
| |
| ' alleged', |
| |
| ' modern', |
| |
| ' Supreme', ' supreme', |
| |
| ' central', |
| |
| ' southern', |
| |
| ' exact', |
| |
| ' core', |
| |
| ' ultimate', ' Ultimate', |
| |
| ' upper', ' Upper', |
| |
| ' mainstream', |
| |
| ' northern', |
| |
| ' remaining', |
| |
| ' immediate', |
| |
| ' eastern', |
| |
| ' western', |
| |
| ' grand', |
| |
| ' Holy', ' holy', |
| |
| ' combined', |
| |
| ' correct', |
| |
| ' underlying', |
| |
| ' proper', |
| |
| ' fundamental', |
| |
| 'rawdownloadcloneembedreportprint', |
| |
| ' formal', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'offic-family', 1 |
|
|
| def prog_L06F08012(input_tokens): |
| """20 token families, 30 surface forms (train F1 of the uncapped optimum: 0.226). |
| Fires at 25,232 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating accumulation or increase""" |
| |
| fire = { |
| |
| ' up', 'up', |
| |
| 'out', ' outs', ' OUT', |
| |
| 'down', ' downs', |
| |
| 'off', |
| |
| 'ass', ' ass', 'assed', |
| |
| 'plus', |
| |
| 'top', 'topic', |
| |
| 'faced', |
| |
| 'door', 'doors', |
| |
| 'looking', 'look', |
| |
| 'Benz', ' Benz', |
| |
| 'new', |
| |
| 'headed', |
| |
| 'world', |
| |
| 'pped', |
| |
| 'handed', |
| |
| 'after', |
| |
| 'cut', |
| |
| 'story', |
| |
| 'hot', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'up-family', 1 |
|
|
| def prog_L06F08013(input_tokens): |
| """50 token families, 52 surface forms (train F1 of the uncapped optimum: 0.668). |
| Fires at 22,747 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to esteemed or high-value opportunities and achievements""" |
| |
| fire = { |
| |
| ' wealthy', |
| |
| ' lengthy', |
| |
| ' expensive', 'expensive', |
| |
| ' prestigious', |
| |
| ' sophisticated', |
| |
| ' ambitious', |
| |
| ' crowded', |
| |
| ' blockbuster', |
| |
| ' lucrative', |
| |
| ' luxury', |
| |
| ' nightclub', |
| |
| ' secretive', |
| |
| ' sprawling', |
| |
| ' coveted', |
| |
| ' hefty', |
| |
| ' hottest', |
| |
| ' elusive', |
| |
| ' dubious', |
| |
| ' prolific', |
| |
| ' damning', |
| |
| ' gruesome', |
| |
| ' hotly', |
| |
| ' undisclosed', |
| |
| ' perks', ' perk', |
| |
| ' sexy', |
| |
| ' booming', |
| |
| ' rigorous', |
| |
| ' tasty', |
| |
| ' prized', |
| |
| ' sizable', |
| |
| ' sensational', |
| |
| ' lavish', |
| |
| ' eagerly', |
| |
| ' bogus', |
| |
| ' handsome', |
| |
| ' marquee', |
| |
| ' juicy', |
| |
| ' upscale', |
| |
| ' slick', |
| |
| ' multim', |
| |
| ' shady', |
| |
| ' trendy', |
| |
| ' affluent', |
| |
| ' unsc', |
| |
| ' murky', |
| |
| ' glamorous', |
| |
| ' stringent', |
| |
| ' plush', |
| |
| ' pricey', |
| |
| ' fashionable', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'wealthy-family', 1 |
|
|
| def prog_L06F08014(input_tokens): |
| """18 token families, 40 surface forms (train F1 of the uncapped optimum: 0.158). |
| Fires at 1,620,205 of 36,786,566 positions. |
| Fires at 4.40% 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 or entities associated with legal or authoritative contexts""" |
| |
| fire = { |
| |
| ',', |
| |
| ' and', 'and', 'anded', ' Anders', |
| |
| '-', ' -', |
| |
| 's', 'S', |
| |
| ' was', 'was', |
| |
| 't', "'t", 'ét', |
| |
| ' (', |
| |
| ' it', |
| |
| ' he', ' He', 'He', |
| |
| ')', ' )', |
| |
| ' not', 'not', ' noting', ' notable', |
| |
| ' said', 'said', |
| |
| ' they', 'they', |
| |
| ' "', |
| |
| ' but', |
| |
| ' can', 'can', |
| |
| ' which', 'which', ' Which', |
| |
| ' what', ' What', 'What', 'what', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F08015(input_tokens): |
| """3 token families, 5 surface forms (train F1 of the uncapped optimum: 0.057). |
| Fires at 80,959 of 36,786,566 positions.""" |
| |
| """prev SAE description: names of players or athletes""" |
| |
| fire = { |
| |
| ',', ' ,', |
| |
| ' (', |
| |
| '),', ' ),', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F08016(input_tokens): |
| """38 token families, 57 surface forms (train F1 of the uncapped optimum: 0.655). |
| Fires at 13,363 of 36,786,566 positions.""" |
| |
| """prev SAE description: activites or phrases suggesting subversion or damage to credibility or authority""" |
| |
| fire = { |
| |
| ' destroy', ' destroying', |
| |
| ' threaten', ' threatens', |
| |
| ' disrupted', ' disrupt', ' disrupting', |
| |
| ' undermining', ' undermined', ' undermines', ' undermin', |
| |
| ' undermine', |
| |
| ' weaken', ' weakened', ' weakening', |
| |
| ' compromised', |
| |
| ' diminished', ' diminish', ' diminishing', |
| |
| ' bolster', |
| |
| ' thwart', ' thwarted', |
| |
| ' betrayed', ' betrayal', ' betray', |
| |
| ' dismantling', ' dismantled', |
| |
| ' derail', ' derailed', |
| |
| ' overthrow', |
| |
| ' uphold', ' upholding', |
| |
| ' discredit', ' discredited', |
| |
| ' overshadowed', ' overshadow', |
| |
| ' dismantle', |
| |
| ' misguided', |
| |
| ' unravel', |
| |
| ' weaker', |
| |
| ' sabotage', |
| |
| ' jeopard', |
| |
| ' circumvent', |
| |
| ' erosion', |
| |
| ' jeopardy', |
| |
| ' obstruct', |
| |
| ' embarrass', |
| |
| ' mockery', |
| |
| ' infiltrate', |
| |
| ' curtail', |
| |
| ' hinder', |
| |
| ' bolstered', |
| |
| ' infiltrated', |
| |
| ' adversaries', |
| |
| ' undercut', |
| |
| ' hampered', |
| |
| ' eroded', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'destroy-family', 1 |
|
|
| def prog_L06F08017(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.491). |
| Fires at 179,499 of 36,786,566 positions. |
| Fires at 0.49% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: instances of the word "in" to highlight locations or contexts""" |
| |
| fire = { |
| |
| ' in', ' IN', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'in-family', 1 |
|
|
| def prog_L06F08018(input_tokens): |
| """3 token families, 5 surface forms (train F1 of the uncapped optimum: 0.829). |
| Fires at 10,330 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the White House""" |
| |
| fire = { |
| |
| ' White', 'White', ' WHITE', |
| |
| ' Downing', |
| |
| ' Whites', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'white-family', 1 |
|
|
| def prog_L06F08019(input_tokens): |
| """25 token families, 32 surface forms (train F1 of the uncapped optimum: 0.161). |
| Fires at 147,203 of 36,786,566 positions. |
| Fires at 0.40% 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: monetary values and price-related information""" |
| |
| fire = { |
| |
| '000', ' 000', |
| |
| ' $', '$', |
| |
| '1', |
| |
| ' million', 'million', ' Million', |
| |
| '2', |
| |
| ' per', |
| |
| ' 18', '18', |
| |
| ' percent', 'percent', |
| |
| '5', |
| |
| '%', |
| |
| ' p', |
| |
| '4', |
| |
| ' billion', |
| |
| '30', |
| |
| '00', |
| |
| ' January', |
| |
| '10', |
| |
| '6', |
| |
| '8', |
| |
| '7', |
| |
| '9', |
| |
| ' cent', ' cents', |
| |
| ' February', |
| |
| '100', |
| |
| ' minutes', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '000-family', 1 |
|
|
| def prog_L06F08020(input_tokens): |
| """13 token families, 61 surface forms (train F1 of the uncapped optimum: 0.243). |
| Fires at 333,939 of 36,786,566 positions. |
| Fires at 0.91% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: words associated with substances and materials""" |
| |
| fire = { |
| |
| ' it', |
| |
| ' data', ' Data', 'Data', 'data', ' DATA', 'DATA', |
| |
| ' money', ' Money', 'Money', 'money', |
| |
| ' drug', ' drugs', ' Drug', 'Drug', ' Drugs', |
| |
| ' car', ' cars', 'Car', 'car', 'cars', ' Cars', |
| |
| ' water', 'water', ' Water', 'Water', |
| |
| ' oil', ' Oil', 'Oil', ' oils', 'oil', |
| |
| ' gun', ' guns', 'Gun', ' Guns', 'guns', ' GUN', |
| |
| ' dog', ' dogs', 'Dog', |
| |
| ' Bitcoin', ' bitcoin', 'Bitcoin', ' bitcoins', ' Bitcoins', 'bitcoin', |
| |
| ' tax', ' taxes', ' Tax', 'Tax', 'tax', ' taxed', ' Taxes', |
| |
| ' beer', ' beers', ' Beer', 'Beer', 'beer', |
| |
| ' food', ' foods', 'Food', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'it-family', 1 |
|
|
| def prog_L06F08021(input_tokens): |
| """24 token families, 33 surface forms (train F1 of the uncapped optimum: 0.747). |
| Fires at 6,815 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "earn" and its variations""" |
| |
| fire = { |
| |
| ' earned', ' earn', ' earning', ' Earn', ' earns', 'earned', 'Earn', |
| |
| ' garnered', |
| |
| ' garner', |
| |
| ' Ear', ' EAR', 'Ear', |
| |
| ' cement', |
| |
| ' nab', |
| |
| ' merit', |
| |
| ' Brave', |
| |
| ' rewarded', |
| |
| 'rob', |
| |
| ' netted', |
| |
| ' clin', |
| |
| ' redeem', ' redeemed', |
| |
| ' incurred', |
| |
| ' laure', |
| |
| ' clinch', |
| |
| ' edged', |
| |
| ' incur', |
| |
| ' accrued', |
| |
| ' forfe', |
| |
| ' reap', |
| |
| 'alien', |
| |
| 'rue', |
| |
| ' leveling', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'earn-family', 1 |
|
|
| def prog_L06F08022(input_tokens): |
| """20 token families, 32 surface forms (train F1 of the uncapped optimum: 0.613). |
| Fires at 11,030 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "Men" in various contexts""" |
| |
| fire = { |
| |
| ' men', ' Men', 'men', 'Men', ' Mens', 'mens', ' MEN', 'MEN', |
| |
| ' Women', 'Women', |
| |
| 'MAN', |
| |
| 'emen', |
| |
| 'femin', ' Femin', |
| |
| 'Med', |
| |
| 'emin', |
| |
| ' Mons', |
| |
| ' Male', 'Male', |
| |
| ' Que', |
| |
| 'Austral', ' Austral', |
| |
| ' menstrual', ' menstru', |
| |
| ' Sally', |
| |
| 'imen', |
| |
| ' meats', |
| |
| ' Mercer', |
| |
| ' ardu', |
| |
| ' sexes', |
| |
| ' menus', |
| |
| ' escal', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'men-family', 1 |
|
|
| def prog_L06F08023(input_tokens): |
| """9 token families, 13 surface forms (train F1 of the uncapped optimum: 0.180). |
| Fires at 59,673 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to sports wins and losses""" |
| |
| fire = { |
| |
| ' over', ' OVER', |
| |
| ' against', |
| |
| ' defeat', ' defeated', ' defeating', |
| |
| ' win', |
| |
| ' loss', |
| |
| ' victory', |
| |
| ' beat', ' beating', |
| |
| ' draw', |
| |
| ' handed', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'over-family', 1 |
|
|
| def prog_L06F08024(input_tokens): |
| """8 token families, 13 surface forms (train F1 of the uncapped optimum: 0.062). |
| Fires at 45,501 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to children's health and related institutions""" |
| |
| fire = { |
| |
| '-', |
| |
| "'", |
| |
| ' children', ' Children', 'children', |
| |
| 'age', ' ages', |
| |
| ' abuse', ' Abuse', 'abuse', |
| |
| 'ic', |
| |
| ' &', |
| |
| ' sex', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '--family', 1 |
|
|
| def prog_L06F08025(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.248). |
| Fires at 76,986 of 36,786,566 positions.""" |
| |
| """prev SAE description: the phrase "standing in" in various contexts""" |
| |
| fire = { |
| |
| ' in', ' IN', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'in-family', 1 |
|
|
| def prog_L06F08026(input_tokens): |
| """25 token families, 65 surface forms (train F1 of the uncapped optimum: 0.317). |
| Fires at 64,103 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to negative impacts or consequences affecting communities or entities""" |
| |
| fire = { |
| |
| ' killed', ' killing', ' kill', ' kills', ' Killed', ' Kills', ' KILL', |
| |
| ' hit', ' hits', 'hit', ' Hits', |
| |
| ' affect', ' affects', ' affecting', ' affected', ' Affect', |
| |
| ' hurt', ' hurts', ' hurting', ' Hurt', |
| |
| ' threatens', ' threaten', ' threatening', |
| |
| ' struck', |
| |
| ' broke', |
| |
| ' shocked', ' shock', ' shocks', 'shock', |
| |
| ' destroyed', ' destroying', ' destroy', ' destroys', 'destroy', |
| |
| ' surrounding', ' surround', ' surrounds', |
| |
| ' facing', |
| |
| ' dominated', ' dominating', ' dominates', |
| |
| ' headlines', ' headlined', |
| |
| ' wounded', ' wounding', |
| |
| ' rocked', ' rocking', |
| |
| ' injuring', |
| |
| ' impacted', ' impacting', ' impacts', |
| |
| ' shaping', ' shaped', |
| |
| ' devastated', ' devastating', |
| |
| ' plagued', |
| |
| ' marked', |
| |
| ' swept', |
| |
| ' hitting', |
| |
| ' haunt', ' haunted', ' haunting', 'Haunted', |
| |
| ' touched', ' touches', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'kill-family', 1 |
|
|
| def prog_L06F08027(input_tokens): |
| """10 token families, 12 surface forms (train F1 of the uncapped optimum: 0.648). |
| Fires at 11,160 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to self and personal agency""" |
| |
| fire = { |
| |
| ' himself', ' Himself', |
| |
| ' themselves', |
| |
| ' itself', |
| |
| ' yourself', ' Yourself', |
| |
| ' myself', |
| |
| ' herself', |
| |
| ' ourselves', |
| |
| 'self', |
| |
| ' oneself', |
| |
| ' yourselves', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'himself-family', 1 |
|
|
| def prog_L06F08028(input_tokens): |
| """47 token families, 59 surface forms (train F1 of the uncapped optimum: 0.705). |
| Fires at 8,429 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to nitrogen and its compounds""" |
| |
| fire = { |
| |
| ' Nit', ' nit', 'Nit', 'nit', |
| |
| ' Liv', 'liv', |
| |
| ' Nec', ' nec', ' NEC', |
| |
| ' Bihar', |
| |
| ' Pure', 'Pure', |
| |
| 'NAT', 'nat', ' NAT', |
| |
| ' Hardy', |
| |
| ' Cumber', |
| |
| ' electr', |
| |
| ' nitrogen', |
| |
| ' NH', |
| |
| ' crit', |
| |
| ' McInt', |
| |
| 'itty', |
| |
| ' coy', |
| |
| ' NV', |
| |
| ' stainless', |
| |
| ' Mit', |
| |
| ' Neh', |
| |
| ' nanop', |
| |
| ' irrit', |
| |
| ' gravel', |
| |
| ' cellul', |
| |
| ' neut', |
| |
| ' Quantum', |
| |
| ' hog', |
| |
| ' nails', |
| |
| 'Nob', 'nob', |
| |
| ' fib', |
| |
| ' oxy', |
| |
| 'NR', 'nr', |
| |
| ' Nig', ' nig', |
| |
| ' Hick', |
| |
| ' Kund', |
| |
| ' Nou', |
| |
| ' Kut', |
| |
| ' Prec', |
| |
| 'ihar', |
| |
| ' neurot', |
| |
| ' partic', |
| |
| ' NG', |
| |
| 'alm', |
| |
| ' Dmit', |
| |
| ' phys', |
| |
| ' Neon', |
| |
| ' pert', |
| |
| ' Nether', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'nit-family', 1 |
|
|
| def prog_L06F08029(input_tokens): |
| """2 token families, 2 surface forms (train F1 of the uncapped optimum: 0.028). |
| Fires at 17,548 of 36,786,566 positions.""" |
| |
| """prev SAE description: statements involving monetary amounts or financial figures""" |
| |
| fire = { |
| |
| '.', |
| |
| 'US', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '.-family', 1 |
|
|
| def prog_L06F08030(input_tokens): |
| """49 token families, 50 surface forms (train F1 of the uncapped optimum: 0.720). |
| Fires at 6,707 of 36,786,566 positions.""" |
| |
| """prev SAE description: the repeated mention of a specific individual named Campbell""" |
| |
| fire = { |
| |
| ' Campbell', |
| |
| 'borough', |
| |
| ' Churchill', |
| |
| 'Camp', |
| |
| ' mascot', |
| |
| ' Davidson', |
| |
| 'coni', |
| |
| ' McMahon', |
| |
| 'bell', |
| |
| ' nostalgia', |
| |
| ' Burlington', |
| |
| ' Canterbury', |
| |
| ' Cunningham', |
| |
| ' Thatcher', |
| |
| ' MacDonald', |
| |
| ' Cannon', |
| |
| 'amoto', |
| |
| ' Burgess', |
| |
| 'imov', |
| |
| 'ingham', |
| |
| ' Cardiff', |
| |
| ' Basil', |
| |
| ' nostalgic', |
| |
| ' Bulgarian', |
| |
| ' Cooke', |
| |
| ' Brooke', |
| |
| ' Ballard', |
| |
| ' Furn', ' furn', |
| |
| 'aughlin', |
| |
| 'uman', |
| |
| ' Townsend', |
| |
| ' modem', |
| |
| ' Barker', |
| |
| ' Chattanooga', |
| |
| 'agher', |
| |
| 'achev', |
| |
| ' salesman', |
| |
| ' resusc', |
| |
| ' counselling', |
| |
| 'rigan', |
| |
| 'umann', |
| |
| ' masc', |
| |
| ' charcoal', |
| |
| '......', |
| |
| ' lantern', |
| |
| ' CAM', |
| |
| ' recru', |
| |
| ' furnace', |
| |
| ' cla', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'campbell-family', 1 |
|
|
| def prog_L06F08031(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.137). |
| Fires at 63,566 of 36,786,566 positions.""" |
| |
| """prev SAE description: the character 'ľ' appearing multiple times""" |
| |
| fire = { |
| |
| '�', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '�-family', 1 |
|
|
| def prog_L06F08032(input_tokens): |
| """39 token families, 47 surface forms (train F1 of the uncapped optimum: 0.569). |
| Fires at 8,791 of 36,786,566 positions.""" |
| |
| """prev SAE description: statements expressing urgency or importance""" |
| |
| fire = { |
| |
| ' highlight', ' highlighting', ' highlighted', |
| |
| ' stressed', ' stresses', ' stressing', |
| |
| ' emphasis', |
| |
| ' stress', 'stress', |
| |
| ' noting', |
| |
| ' emphasized', ' emphasizes', |
| |
| ' reiterated', |
| |
| ' neglected', |
| |
| ' emphasize', |
| |
| ' mantra', |
| |
| ' cautioned', |
| |
| ' convey', ' conveyed', |
| |
| ' undersc', |
| |
| ' underscores', |
| |
| ' preach', ' preached', |
| |
| ' emphas', |
| |
| ' emphasizing', |
| |
| ' hammered', |
| |
| ' underscore', |
| |
| ' touting', |
| |
| 'lined', |
| |
| ' reiterate', |
| |
| ' clarified', |
| |
| ' reinforce', |
| |
| ' reiter', |
| |
| ' Nay', |
| |
| ' simplified', |
| |
| 'phas', |
| |
| ' emphatically', |
| |
| ' reinforces', |
| |
| ' exagger', |
| |
| ' prioritize', |
| |
| ' affirm', |
| |
| ' sweating', |
| |
| ' gloss', |
| |
| 'vich', |
| |
| 'ificantly', |
| |
| ' emph', |
| |
| 'cially', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'highlight-family', 1 |
|
|
| def prog_L06F08033(input_tokens): |
| """25 token families, 34 surface forms (train F1 of the uncapped optimum: 0.552). |
| Fires at 9,836 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to constitutional amendments, particularly the Second and Fourth Amendments""" |
| |
| fire = { |
| |
| ' Second', 'Second', 'second', ' Secondly', ' Seconds', ' secondly', 'Secondly', |
| |
| ' First', |
| |
| ' Third', 'Third', |
| |
| ' Fourth', 'Fourth', 'fourth', |
| |
| ' Fifth', |
| |
| ' Latter', |
| |
| 'ennial', |
| |
| ' Sixth', |
| |
| ' Seventh', |
| |
| ' Ninth', |
| |
| ' Constitutional', |
| |
| ' Inner', |
| |
| ' Moment', |
| |
| 'ь', |
| |
| ' Eighth', |
| |
| ' Revised', |
| |
| ' Tenth', |
| |
| 'venth', |
| |
| 'Fif', |
| |
| 'Iss', |
| |
| ' twentieth', |
| |
| ' Occ', |
| |
| 'elfth', |
| |
| 'ieth', |
| |
| ' Semin', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'second-family', 1 |
|
|
| def prog_L06F08034(input_tokens): |
| """1 token families, 2 surface forms (train F1 of the uncapped optimum: 0.599). |
| Fires at 91,617 of 36,786,566 positions.""" |
| |
| """prev SAE description: repeated references to the word "it." """ |
| |
| fire = { |
| |
| ' it', ' It', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'it-family', 1 |
|
|
| def prog_L06F08035(input_tokens): |
| """12 token families, 15 surface forms (train F1 of the uncapped optimum: 0.112). |
| Fires at 22,908 of 36,786,566 positions.""" |
| |
| """prev SAE description: attributions or copyright statements""" |
| |
| fire = { |
| |
| ' by', ' By', ' BY', 'BY', |
| |
| ' All', |
| |
| ' Facebook', |
| |
| 'IS', |
| |
| 'Published', |
| |
| 'ING', |
| |
| ' supported', |
| |
| 'OU', |
| |
| ' expressed', |
| |
| ' Picks', |
| |
| ' 2019', |
| |
| 'rop', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'by-family', 1 |
|
|
| def prog_L06F08036(input_tokens): |
| """16 token families, 21 surface forms (train F1 of the uncapped optimum: 0.250). |
| Fires at 22,492 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases indicating actions taken or steps proposed in various contexts""" |
| |
| fire = { |
| |
| ' steps', ' step', ' Steps', |
| |
| ' efforts', ' effort', |
| |
| ' action', |
| |
| ' measures', |
| |
| ' progress', |
| |
| ' stepped', ' stepping', |
| |
| ' aim', |
| |
| ' moves', |
| |
| ' unanimously', |
| |
| ' strides', |
| |
| ' momentum', |
| |
| ' leap', ' leaps', |
| |
| ' groundwork', |
| |
| ' stance', |
| |
| ' lengths', |
| |
| ' oath', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'step-family', 1 |
|
|
| def prog_L06F08037(input_tokens): |
| """22 token families, 29 surface forms (train F1 of the uncapped optimum: 0.879). |
| Fires at 7,061 of 36,786,566 positions.""" |
| |
| """prev SAE description: occurrences of the name "Tom." """ |
| |
| fire = { |
| |
| ' Tom', 'Tom', ' tom', 'tom', ' TOM', |
| |
| ' Tomb', ' tomb', |
| |
| ' Kit', |
| |
| ' Dom', |
| |
| ' Newt', |
| |
| ' Thom', 'Thom', |
| |
| ' Nic', |
| |
| ' tub', |
| |
| ' Uncle', |
| |
| ' Tomas', |
| |
| ' Kot', |
| |
| ' Ton', 'Ton', |
| |
| ' Chomsky', |
| |
| ' Nom', |
| |
| ' tun', |
| |
| ' knowingly', |
| |
| ' tart', |
| |
| ' Tad', |
| |
| ' esoteric', |
| |
| ' torped', |
| |
| ' Tome', |
| |
| ' soph', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'tom-family', 1 |
|
|
| def prog_L06F08038(input_tokens): |
| """25 token families, 34 surface forms (train F1 of the uncapped optimum: 0.655). |
| Fires at 8,076 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to boats""" |
| |
| fire = { |
| |
| ' boat', ' boats', 'boat', ' Boat', 'boats', |
| |
| ' vessel', ' vessels', |
| |
| ' helicopter', |
| |
| ' ferry', ' Ferry', |
| |
| ' sail', ' sailed', ' sails', |
| |
| ' knot', ' knots', |
| |
| ' raft', |
| |
| 'acht', |
| |
| ' kay', |
| |
| ' porch', |
| |
| ' balcony', |
| |
| 'oat', |
| |
| ' yacht', |
| |
| ' padd', |
| |
| ' canoe', |
| |
| 'arge', |
| |
| ' RV', |
| |
| ' dock', |
| |
| ' ding', |
| |
| ' ashore', |
| |
| ' paddle', |
| |
| ' buoy', |
| |
| ' overboard', |
| |
| ' fisherman', |
| |
| ' baskets', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'boat-family', 1 |
|
|
| def prog_L06F08039(input_tokens): |
| """1 token families, 3 surface forms (train F1 of the uncapped optimum: 0.416). |
| Fires at 26,920 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "that" in various contexts""" |
| |
| fire = { |
| |
| ' That', 'That', 'that', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'that-family', 1 |
|
|
| def prog_L06F08040(input_tokens): |
| """41 token families, 51 surface forms (train F1 of the uncapped optimum: 0.689). |
| Fires at 5,797 of 36,786,566 positions.""" |
| |
| """prev SAE description: the term "dispense" along with related vocabulary indicating distribution or provision""" |
| |
| fire = { |
| |
| ' dispatch', ' dispatched', ' dispatcher', |
| |
| 'Resp', ' resp', ' Resp', ' RESP', |
| |
| ' disp', ' Disp', |
| |
| 'Fl', |
| |
| ' readily', |
| |
| ' diagn', ' Diagn', 'diagn', |
| |
| ' Pend', ' pend', |
| |
| ' inqu', |
| |
| ' Prec', |
| |
| ' desp', |
| |
| ' dispens', |
| |
| ' dup', |
| |
| ' discontin', |
| |
| ' sectarian', |
| |
| ' inund', |
| |
| ' dissemin', |
| |
| ' juxtap', |
| |
| ' decon', |
| |
| ' pepp', |
| |
| ' condesc', |
| |
| ' fam', ' FAM', |
| |
| ' bureaucr', |
| |
| ' disinteg', |
| |
| ' Pharm', |
| |
| ' confisc', |
| |
| ' Coy', |
| |
| ' eman', |
| |
| ' propriet', |
| |
| ' deline', |
| |
| ' expl', |
| |
| ' flatt', |
| |
| ' annihil', |
| |
| ' brist', |
| |
| ' popul', |
| |
| ' promulg', |
| |
| ' COMP', |
| |
| ' bart', |
| |
| ' retri', |
| |
| 'REP', |
| |
| ' Millenn', |
| |
| ' intim', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'dispatch-family', 1 |
|
|
| def prog_L06F08041(input_tokens): |
| """2 token families, 3 surface forms (train F1 of the uncapped optimum: 0.040). |
| Fires at 21,839 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases that question the validity or authenticity of statements or situations""" |
| |
| fire = { |
| |
| ' there', ' THERE', |
| |
| ' anyone', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'there-family', 1 |
|
|
| def prog_L06F08042(input_tokens): |
| """50 token families, 52 surface forms (train F1 of the uncapped optimum: 0.602). |
| Fires at 12,520 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the concept of invisibility""" |
| |
| fire = { |
| |
| 'able', |
| |
| ' invisible', ' Invisible', |
| |
| ' unnecessary', |
| |
| ' Jackie', |
| |
| ' elusive', |
| |
| ' anonymous', |
| |
| ' injustice', |
| |
| ' unnoticed', |
| |
| ' immortal', ' Immortal', |
| |
| 'asha', |
| |
| ' unseen', |
| |
| ' imaginary', |
| |
| 'ibility', |
| |
| ' Ig', |
| |
| ' unconscious', |
| |
| ' airborne', |
| |
| ' unstoppable', |
| |
| ' ineffective', |
| |
| ' useless', |
| |
| ' pesky', |
| |
| ' horizontal', |
| |
| ' ghosts', |
| |
| ' Jonas', |
| |
| ' hem', |
| |
| ' Wilkinson', |
| |
| ' lev', |
| |
| ' unpaid', |
| |
| ' irrelevant', |
| |
| 'Untitled', |
| |
| ' inaccessible', |
| |
| ' Stevenson', |
| |
| ' underdog', |
| |
| ' Heidi', |
| |
| ' involuntary', |
| |
| ' underestimate', |
| |
| ' irresistible', |
| |
| 'assuming', |
| |
| ' nonexistent', |
| |
| ' powerless', |
| |
| ' Sylvia', |
| |
| ' electromagnetic', |
| |
| ' oblivious', |
| |
| ' Kitty', |
| |
| ' opaque', |
| |
| ' undet', |
| |
| ' inactive', |
| |
| 'ionic', |
| |
| ' angels', |
| |
| 'visible', |
| |
| ' Omn', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'able-family', 1 |
|
|
| def prog_L06F08043(input_tokens): |
| """34 token families, 46 surface forms (train F1 of the uncapped optimum: 0.279). |
| Fires at 47,445 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to potential candidates or selections in various contexts""" |
| |
| fire = { |
| |
| ' choice', ' Choice', 'choice', |
| |
| ' candidate', ' Candidate', |
| |
| ' bid', |
| |
| ' effort', |
| |
| ' attempt', |
| |
| ' answer', |
| |
| ' option', |
| |
| ' chance', |
| |
| ' favorite', 'favorite', |
| |
| ' candidates', |
| |
| ' contender', ' contenders', |
| |
| ' target', |
| |
| ' winner', |
| |
| ' threat', |
| |
| ' nominee', |
| |
| ' push', |
| |
| ' entry', |
| |
| ' convinced', ' convincing', |
| |
| ' favourite', |
| |
| ' bet', ' bets', |
| |
| ' priority', 'priority', |
| |
| ' selection', |
| |
| ' boost', |
| |
| 'runner', 'runners', |
| |
| ' favorites', |
| |
| ' eligible', 'eligible', |
| |
| ' choices', |
| |
| ' chances', |
| |
| ' unlikely', |
| |
| ' quest', |
| |
| ' appointment', |
| |
| ' qualified', 'qualified', ' qualifies', |
| |
| ' finalists', |
| |
| ' culprit', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'choice-family', 1 |
|
|
| def prog_L06F08044(input_tokens): |
| """3 token families, 5 surface forms (train F1 of the uncapped optimum: 0.337). |
| Fires at 8,693 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases emphasizing uniqueness or the exclusive nature of something""" |
| |
| fire = { |
| |
| ' only', 'only', ' ONLY', |
| |
| ' sole', |
| |
| ' lone', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'only-family', 1 |
|
|
| def prog_L06F08045(input_tokens): |
| """3 token families, 4 surface forms (train F1 of the uncapped optimum: 0.958). |
| Fires at 4,160 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of individuals named George, particularly former presidents""" |
| |
| fire = { |
| |
| ' George', 'George', |
| |
| ' Osama', |
| |
| 'ORGE', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'george-family', 1 |
|
|
| def prog_L06F08046(input_tokens): |
| """45 token families, 49 surface forms (train F1 of the uncapped optimum: 0.696). |
| Fires at 8,435 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the name "Adams." """ |
| |
| fire = { |
| |
| ' Adams', 'Adams', |
| |
| ' Rodgers', |
| |
| ' Abrams', ' Abram', |
| |
| ' Jenkins', |
| |
| 'West', |
| |
| ' Pak', 'Pak', |
| |
| ' outdoors', |
| |
| ' Abbas', |
| |
| ' Jennings', |
| |
| ' Andrews', |
| |
| ' Independence', |
| |
| ' Barton', |
| |
| ' Berg', |
| |
| ' Yates', |
| |
| ' Bates', |
| |
| ' Beaver', |
| |
| 'Â', |
| |
| ' Pike', |
| |
| ' independents', |
| |
| ' Daniels', |
| |
| ' Hawkins', |
| |
| ' Prometheus', |
| |
| ' Hancock', |
| |
| 'monds', |
| |
| ' Sag', |
| |
| ' colonies', |
| |
| ' Amer', 'Americ', |
| |
| ' Goff', |
| |
| ' Townsend', |
| |
| ' pigeon', |
| |
| ' Dickinson', |
| |
| ' Augusta', |
| |
| ' hops', |
| |
| ' Aliens', |
| |
| ' Ballard', |
| |
| ' Albion', |
| |
| ' Oswald', |
| |
| ' Bahamas', |
| |
| ' Ö', |
| |
| ' inert', |
| |
| ' Pax', |
| |
| ' alienation', |
| |
| ' Ü', |
| |
| 'AMS', |
| |
| ' colonists', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'adam-family', 1 |
|
|
| def prog_L06F08047(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.071). |
| Fires at 50,079 of 36,786,566 positions.""" |
| |
| """prev SAE description: the phrase "that being said." """ |
| |
| fire = { |
| |
| ',', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F08048(input_tokens): |
| """18 token families, 27 surface forms (train F1 of the uncapped optimum: 0.833). |
| Fires at 6,325 of 36,786,566 positions.""" |
| |
| """prev SAE description: repeated mentions of the suffix "af" """ |
| |
| fire = { |
| |
| 'af', 'AF', 'Af', ' Af', ' af', ' AF', |
| |
| 'uf', |
| |
| ' Saf', ' SAF', ' saf', 'saf', |
| |
| ' Raf', 'raf', |
| |
| 'afa', |
| |
| 'afe', |
| |
| 'afi', |
| |
| ' Haf', |
| |
| 'afer', |
| |
| 'CF', |
| |
| 'Afee', |
| |
| ' Sof', |
| |
| ' McF', |
| |
| ' Kaf', |
| |
| ' FreeBSD', |
| |
| ' Caf', |
| |
| 'afia', |
| |
| ' Laf', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'af-family', 1 |
|
|
| def prog_L06F08049(input_tokens): |
| """45 token families, 56 surface forms (train F1 of the uncapped optimum: 0.719). |
| Fires at 10,740 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to zombies and zombie-related themes in various contexts""" |
| |
| fire = { |
| |
| ' zombie', ' Zombie', |
| |
| ' Dead', |
| |
| ' Vampire', ' vampire', |
| |
| ' skeleton', ' skeletons', |
| |
| ' zombies', ' Zombies', |
| |
| ' Ninja', ' ninja', |
| |
| ' AMC', |
| |
| ' Mutant', ' mutant', ' mutants', |
| |
| ' newborn', |
| |
| ' apocalypse', |
| |
| ' Giant', |
| |
| ' livestock', |
| |
| ' communist', |
| |
| ' Romero', |
| |
| ' uninsured', |
| |
| ' Resident', |
| |
| 'ocalyptic', |
| |
| ' Chern', |
| |
| ' Deadpool', |
| |
| ' corpse', ' Corpse', |
| |
| ' Zhang', |
| |
| ' GMO', |
| |
| 'iem', |
| |
| ' undead', ' Undead', |
| |
| ' gn', |
| |
| ' corpses', |
| |
| ' Marxist', |
| |
| 'cow', 'Cow', |
| |
| 'arant', |
| |
| ' Savannah', |
| |
| ' hordes', |
| |
| ' reim', |
| |
| ' apocalyptic', |
| |
| ' Zy', |
| |
| 'ummies', |
| |
| ' quarantine', |
| |
| ' Aliens', |
| |
| ' Slayer', |
| |
| ' horde', ' Horde', |
| |
| 'ombie', |
| |
| ' feral', |
| |
| ' Garg', |
| |
| ' vampires', |
| |
| ' avatar', |
| |
| 'oodoo', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'zombie-family', 1 |
|
|
| def prog_L06F08050(input_tokens): |
| """46 token families, 58 surface forms (train F1 of the uncapped optimum: 0.671). |
| Fires at 11,273 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to jungles and swamps""" |
| |
| fire = { |
| |
| ' Sur', |
| |
| ' jungle', ' Jungle', |
| |
| ' backpack', |
| |
| ' tropical', ' Tropical', |
| |
| ' Paradise', ' paradise', |
| |
| ' queue', 'queue', 'Queue', |
| |
| ' bush', |
| |
| ' Jung', ' jung', |
| |
| ' swamp', ' Swamp', |
| |
| ' Jurassic', |
| |
| 'jin', |
| |
| ' Wilderness', ' wilderness', |
| |
| ' Calais', |
| |
| 'forest', |
| |
| ' Congo', |
| |
| ' Beats', |
| |
| ' ambush', |
| |
| ' Hun', |
| |
| ' DK', 'DK', 'dk', |
| |
| ' hind', |
| |
| ' camouflage', |
| |
| ' Jade', |
| |
| 'jay', |
| |
| ' trou', |
| |
| ' Kab', |
| |
| ' dystopian', |
| |
| ' Sutton', |
| |
| ' transitional', |
| |
| ' nursery', |
| |
| ' Nightmare', |
| |
| ' lush', |
| |
| 'Jump', |
| |
| ' bamboo', |
| |
| ' backdoor', |
| |
| ' Razer', |
| |
| ' attic', |
| |
| ' Gap', |
| |
| ' Sahara', |
| |
| ' slime', ' Slime', |
| |
| ' Gorge', |
| |
| ' mars', 'mares', |
| |
| ' queues', |
| |
| ' ding', |
| |
| 'oodoo', |
| |
| 'ungle', |
| |
| ' Junk', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'sur-family', 1 |
|
|
| def prog_L06F08051(input_tokens): |
| """9 token families, 11 surface forms (train F1 of the uncapped optimum: 0.556). |
| Fires at 20,415 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "called" referring to definitions or labels""" |
| |
| fire = { |
| |
| ' called', ' Called', ' CALL', |
| |
| ' referred', |
| |
| ' dubbed', |
| |
| ' nicknamed', |
| |
| ' labeled', |
| |
| 'itled', |
| |
| ' renamed', |
| |
| ' termed', |
| |
| ' labelled', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'call-family', 1 |
|
|
| def prog_L06F08052(input_tokens): |
| """1 token families, 1 surface forms (train F1 of the uncapped optimum: 0.148). |
| Fires at 72,214 of 36,786,566 positions.""" |
| |
| """prev SAE description: repeated instances of a specific symbol or character sequence""" |
| |
| fire = { |
| |
| '�', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '�-family', 1 |
|
|
| def prog_L06F08053(input_tokens): |
| """24 token families, 26 surface forms (train F1 of the uncapped optimum: 0.575). |
| Fires at 7,060 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to legal charges or indictments""" |
| |
| fire = { |
| |
| ' indictment', ' indicted', ' indict', |
| |
| ' defendants', |
| |
| ' wiret', |
| |
| ' impeachment', |
| |
| ' extradition', |
| |
| ' unse', |
| |
| ' genus', |
| |
| ' arra', |
| |
| ' autopsy', |
| |
| ' damning', |
| |
| ' antitrust', |
| |
| ' Inspector', |
| |
| ' conspiring', |
| |
| ' extrad', |
| |
| 'Prosecutors', |
| |
| ' criminally', |
| |
| ' implicated', |
| |
| ' recommending', |
| |
| 'eering', |
| |
| ' prosecute', |
| |
| ' incrim', |
| |
| ' acqu', |
| |
| 'achment', |
| |
| 'ossier', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'indict-family', 1 |
|
|
| def prog_L06F08054(input_tokens): |
| """5 token families, 9 surface forms (train F1 of the uncapped optimum: 0.338). |
| Fires at 74,082 of 36,786,566 positions.""" |
| |
| """prev SAE description: possessive pronouns and their frequency""" |
| |
| fire = { |
| |
| ' his', |
| |
| ' their', ' THEIR', |
| |
| ' its', |
| |
| ' your', ' Your', 'your', ' YOUR', |
| |
| ' my', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'his-family', 1 |
|
|
| def prog_L06F08055(input_tokens): |
| """29 token families, 32 surface forms (train F1 of the uncapped optimum: 0.835). |
| Fires at 7,029 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to the name "Nick." """ |
| |
| fire = { |
| |
| ' Nick', 'Nick', 'nick', ' nick', |
| |
| ' boyfriend', |
| |
| 'Jeff', |
| |
| ' Nicholas', |
| |
| ' Nate', |
| |
| 'nik', |
| |
| ' Diaz', |
| |
| ' Ibrahim', |
| |
| 'Jud', |
| |
| ' nickname', |
| |
| ' Schmidt', |
| |
| ' fox', |
| |
| 'ICK', |
| |
| ' Nickel', |
| |
| ' Judy', |
| |
| ' Griffith', |
| |
| ' Jess', |
| |
| ' Nad', |
| |
| 'Nich', |
| |
| 'Beck', |
| |
| ' pup', |
| |
| ' Ned', |
| |
| ' Norm', |
| |
| 'Buzz', |
| |
| ' bunny', |
| |
| 'onym', |
| |
| 'midt', |
| |
| ' glyph', |
| |
| ' mism', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'nick-family', 1 |
|
|
| def prog_L06F08056(input_tokens): |
| """39 token families, 71 surface forms (train F1 of the uncapped optimum: 0.462). |
| Fires at 77,821 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of conflict or fighting""" |
| |
| fire = { |
| |
| ' fight', ' fighting', ' fights', 'fight', ' Fight', 'fighting', 'Fighting', 'fights', 'Fight', ' FIGHT', |
| |
| ' war', ' wars', 'Wars', |
| |
| ' argument', ' argued', ' arguments', ' arguing', ' Argument', |
| |
| ' battle', 'battle', |
| |
| ' debate', ' Debate', |
| |
| ' match', ' Match', 'match', |
| |
| ' clashes', ' clash', ' clashed', |
| |
| ' conversation', |
| |
| ' battles', ' battling', ' battled', ' Battles', |
| |
| ' discussion', 'Discuss', |
| |
| ' conflict', ' conflicts', |
| |
| ' dispute', |
| |
| ' relationship', |
| |
| ' row', |
| |
| ' exchange', |
| |
| ' argue', |
| |
| ' struggle', ' Struggle', |
| |
| ' negotiations', |
| |
| ' discussions', |
| |
| ' fought', |
| |
| ' encounter', ' encounters', |
| |
| ' contest', |
| |
| ' heated', |
| |
| ' confrontation', |
| |
| ' riots', ' riot', |
| |
| ' standoff', |
| |
| ' chase', |
| |
| ' tensions', |
| |
| ' bout', ' bouts', |
| |
| ' showdown', ' Showdown', |
| |
| ' rivalry', |
| |
| ' feud', |
| |
| ' collision', |
| |
| ' gunfire', |
| |
| ' contentious', |
| |
| ' tension', |
| |
| ' rant', |
| |
| ' brawl', ' Brawl', |
| |
| ' altercation', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'fight-family', 1 |
|
|
| def prog_L06F08057(input_tokens): |
| """12 token families, 60 surface forms (train F1 of the uncapped optimum: 0.205). |
| Fires at 474,837 of 36,786,566 positions. |
| Fires at 1.29% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: significant events or actions related to conflict and struggle""" |
| |
| fire = { |
| |
| ' year', ' Year', ' yearly', |
| |
| ' season', 'season', ' seasons', ' Season', ' seasonal', 'Season', ' seasoning', |
| |
| ' support', ' supporters', ' supportive', 'support', ' SUPPORT', |
| |
| ' life', ' LIFE', |
| |
| ' campaign', ' campaigns', ' Campaign', ' campaigning', 'campaign', |
| |
| ' record', ' records', ' recording', ' recordings', 'record', ' Record', ' Recording', ' recorder', 'recorded', |
| |
| ' play', ' playing', 'play', ' playable', 'playing', 'Play', ' PLAY', |
| |
| ' death', ' deaths', 'death', |
| |
| ' game', |
| |
| ' attack', ' attacks', 'attack', 'attacks', |
| |
| ' deal', ' deals', ' dealings', ' Deals', ' Deal', 'deal', 'Deal', |
| |
| ' event', ' events', ' Event', ' Events', 'Event', 'events', ' EVENT', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'year-family', 1 |
|
|
| def prog_L06F08058(input_tokens): |
| """26 token families, 37 surface forms (train F1 of the uncapped optimum: 0.662). |
| Fires at 7,922 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to contributions or factors leading to an outcome or effect""" |
| |
| fire = { |
| |
| ' contributed', ' contributing', ' Contribut', ' contributes', 'Contribut', ' contribut', |
| |
| ' contributor', ' contributors', |
| |
| ' contribute', |
| |
| ' contribution', |
| |
| ' donated', ' donating', |
| |
| ' contributions', ' Contributions', |
| |
| 'ributed', 'ribut', |
| |
| 'authored', ' authored', |
| |
| ' owing', |
| |
| ' Produ', |
| |
| 'ribute', |
| |
| ' yielded', ' yielding', |
| |
| ' COM', |
| |
| ' Jamal', |
| |
| ' dictate', |
| |
| ' detrimental', |
| |
| ' dictated', |
| |
| ' collaborated', |
| |
| ' succumbed', |
| |
| ' detract', |
| |
| ' attributable', |
| |
| 'MENTS', |
| |
| 'ribution', |
| |
| '________________________________', |
| |
| ' CONTR', |
| |
| ' aggrav', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'contribut-family', 1 |
|
|
| def prog_L06F08059(input_tokens): |
| """21 token families, 21 surface forms (train F1 of the uncapped optimum: 0.848). |
| Fires at 3,235 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the name "Muhammad." """ |
| |
| fire = { |
| |
| ' Mohammed', |
| |
| ' Muhammad', |
| |
| ' Mohammad', |
| |
| ' Noble', |
| |
| ' Moses', |
| |
| ' Medina', |
| |
| ' Prophet', |
| |
| ' Allah', |
| |
| ' Messenger', |
| |
| ' Mustang', |
| |
| 'Mu', |
| |
| ' Ferdinand', |
| |
| ' Lucy', |
| |
| ' Frazier', |
| |
| ' Mecca', |
| |
| ' Spock', |
| |
| ' Emerson', |
| |
| 'ammad', |
| |
| 'hammad', |
| |
| '.......', |
| |
| 'dinand', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'mohamm-family', 1 |
|
|
| def prog_L06F08060(input_tokens): |
| """23 token families, 47 surface forms (train F1 of the uncapped optimum: 0.146). |
| Fires at 521,601 of 36,786,566 positions. |
| Fires at 1.42% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: phrases indicating plans or intentions""" |
| |
| fire = { |
| |
| '\n', |
| |
| ' I', 'I', |
| |
| ' The', |
| |
| ' we', ' We', |
| |
| ' you', ' You', 'You', |
| |
| ' This', 'This', |
| |
| ' It', |
| |
| ' our', ' Our', 'Our', |
| |
| ' also', ' Also', 'Also', |
| |
| ' some', ' Some', 'Some', |
| |
| ' how', 'How', ' How', |
| |
| ' my', ' My', 'My', |
| |
| ' In', |
| |
| ' If', 'If', |
| |
| '!', |
| |
| ' But', |
| |
| ' And', 'And', |
| |
| ' For', 'For', |
| |
| ' There', 'There', |
| |
| ' Here', 'Here', |
| |
| ' What', 'What', 'what', |
| |
| ' That', 'That', |
| |
| ' So', 'So', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '\\n-family', 1 |
|
|
| def prog_L06F08061(input_tokens): |
| """8 token families, 13 surface forms (train F1 of the uncapped optimum: 0.114). |
| Fires at 34,267 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to social sciences and general surveys""" |
| |
| fire = { |
| |
| ' manager', ' Manager', ' managers', |
| |
| ' election', ' Election', |
| |
| ' Assembly', ' assembly', |
| |
| ' Motors', |
| |
| ' Hospital', |
| |
| ' Catholic', |
| |
| ' counsel', ' Counsel', |
| |
| ' bargaining', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'manag-family', 1 |
|
|
| def prog_L06F08062(input_tokens): |
| """49 token families, 54 surface forms (train F1 of the uncapped optimum: 0.652). |
| Fires at 17,516 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to envelopes and related terms""" |
| |
| fire = { |
| |
| ' SUV', |
| |
| ' Erik', |
| |
| ' aw', |
| |
| ' armored', ' Armored', |
| |
| ' engulfed', ' engulf', |
| |
| ' stuffed', |
| |
| ' purse', |
| |
| ' Airbus', |
| |
| ' Agg', 'Agg', |
| |
| ' SWAT', |
| |
| ' Naj', |
| |
| ' Wald', |
| |
| ' Oversight', |
| |
| ' nephew', |
| |
| ' envelope', |
| |
| 'ENG', |
| |
| ' elegant', |
| |
| ' Monetary', |
| |
| ' gripped', ' gripping', |
| |
| ' envelop', |
| |
| ' Excel', |
| |
| ' shootout', |
| |
| 'advertisement', |
| |
| ' Flores', |
| |
| ' XV', |
| |
| ' Won', |
| |
| ' clut', |
| |
| ' Fuj', |
| |
| ' shrouded', ' Shroud', |
| |
| 'lyss', |
| |
| ' Ek', |
| |
| ' Cay', |
| |
| ' unab', |
| |
| ' Keys', |
| |
| ' playbook', |
| |
| ' Westbrook', |
| |
| ' perplex', |
| |
| ' mountainous', |
| |
| 'velop', |
| |
| ' FedEx', |
| |
| ' Huma', |
| |
| ' EUR', |
| |
| ' armoured', |
| |
| ' oversized', |
| |
| ' Edwin', |
| |
| ' flo', |
| |
| ' enclosed', |
| |
| ' purs', |
| |
| ' Deng', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'suv-family', 1 |
|
|
| def prog_L06F08063(input_tokens): |
| """19 token families, 30 surface forms (train F1 of the uncapped optimum: 0.144). |
| Fires at 56,764 of 36,786,566 positions.""" |
| |
| """prev SAE description: words or phrases indicating specific actions or conditions""" |
| |
| fire = { |
| |
| 'on', 'ON', |
| |
| 'a', |
| |
| 'i', |
| |
| 'ar', |
| |
| 'ro', 'RO', |
| |
| 'own', 'OWN', |
| |
| 'el', |
| |
| 'te', 'TE', |
| |
| 'TA', |
| |
| 'ri', 'RI', |
| |
| 'im', 'IM', |
| |
| 'ah', 'AH', |
| |
| 'ra', |
| |
| 'ale', 'ALE', |
| |
| 'iv', 'IV', |
| |
| 'ime', 'imes', |
| |
| 'ron', 'ronic', |
| |
| 'RE', |
| |
| 'ore', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'on-family', 1 |
|
|
| def prog_L06F08064(input_tokens): |
| """15 token families, 23 surface forms (train F1 of the uncapped optimum: 0.092). |
| Fires at 513,448 of 36,786,566 positions. |
| Fires at 1.40% 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: indications of audience satisfaction and reactions""" |
| |
| fire = { |
| |
| ' and', ' And', |
| |
| ' that', |
| |
| ' it', ' It', |
| |
| ' I', |
| |
| ' but', ' But', 'but', |
| |
| ' they', ' They', |
| |
| ' just', ' Just', 'just', |
| |
| ' about', |
| |
| ' my', |
| |
| ' what', |
| |
| ' there', 'thereal', |
| |
| ' some', |
| |
| ' time', |
| |
| ' how', |
| |
| ' like', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'and-family', 1 |
|
|
| def prog_L06F08065(input_tokens): |
| """29 token families, 32 surface forms (train F1 of the uncapped optimum: 0.346). |
| Fires at 55,909 of 36,786,566 positions.""" |
| |
| """prev SAE description: connective phrases that introduce comparisons or conditions""" |
| |
| fire = { |
| |
| 'the', |
| |
| 'or', |
| |
| 'which', |
| |
| 'not', |
| |
| 'including', 'includes', |
| |
| 'with', |
| |
| 'for', |
| |
| 'who', 'WHO', |
| |
| 'born', |
| |
| 'also', |
| |
| 'like', 'likely', |
| |
| 'that', |
| |
| 'see', |
| |
| 'from', |
| |
| 'via', |
| |
| 'pictured', |
| |
| 'but', |
| |
| 'though', |
| |
| 'left', |
| |
| 'even', |
| |
| 'now', |
| |
| 'aka', |
| |
| 'such', |
| |
| 'especially', |
| |
| 'although', |
| |
| 'where', |
| |
| 'yes', |
| |
| 'about', |
| |
| 'when', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'the-family', 1 |
|
|
| def prog_L06F08066(input_tokens): |
| """49 token families, 54 surface forms (train F1 of the uncapped optimum: 0.743). |
| Fires at 12,325 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the name "Bryant." """ |
| |
| fire = { |
| |
| ' Lakers', |
| |
| ' Bryant', |
| |
| ' Everton', |
| |
| 'lon', ' Lon', |
| |
| ' Brent', |
| |
| 'Brien', |
| |
| 'obe', |
| |
| ' Ibrahim', |
| |
| ' Beverly', |
| |
| ' Cardiff', |
| |
| ' Dwight', |
| |
| 'arks', |
| |
| ' Baylor', |
| |
| 'obi', ' Obi', |
| |
| ' BART', |
| |
| 'oby', |
| |
| ' Bry', |
| |
| ' Bolton', |
| |
| ' Kobe', |
| |
| ' Lionel', |
| |
| ' Walton', |
| |
| ' Dol', |
| |
| ' Rodney', |
| |
| ' Byron', |
| |
| ' Marks', |
| |
| ' Kardashian', |
| |
| ' leveled', ' leveling', |
| |
| ' Archer', |
| |
| ' rubbing', ' rubbed', |
| |
| ' stub', ' Stub', |
| |
| ' Staples', |
| |
| ' Sears', |
| |
| ' Pau', |
| |
| 'seys', |
| |
| ' Ingram', |
| |
| 'nard', |
| |
| ' Barkley', |
| |
| ' wardrobe', |
| |
| ' singled', |
| |
| 'obyl', |
| |
| ' disarm', |
| |
| ' Chandra', |
| |
| ' tuberculosis', |
| |
| ' Kry', |
| |
| ' Chamberlain', |
| |
| ' Payton', |
| |
| ' Kare', |
| |
| ' Wilde', |
| |
| ' Ramsay', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'lak-family', 1 |
|
|
| def prog_L06F08067(input_tokens): |
| """42 token families, 58 surface forms (train F1 of the uncapped optimum: 0.701). |
| Fires at 9,497 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the striatum in neuronal contexts""" |
| |
| fire = { |
| |
| 'RI', |
| |
| ' Stro', 'stro', ' stro', |
| |
| ' stri', 'stri', ' Stri', |
| |
| 'Dri', ' Dri', 'dri', |
| |
| ' trou', ' Trou', |
| |
| ' Wim', |
| |
| ' Sioux', |
| |
| ' Chall', 'Chall', 'chall', ' chall', |
| |
| 'stru', ' Stru', |
| |
| ' Charg', 'Charg', |
| |
| 'pri', 'Pri', |
| |
| ' Entreprene', |
| |
| ' PRO', 'PRO', |
| |
| ' Haj', |
| |
| ' Stam', |
| |
| ' Pau', |
| |
| ' Pes', |
| |
| ' Cra', 'Cra', |
| |
| ' scra', |
| |
| ' chast', |
| |
| ' flo', 'flo', |
| |
| ' Wide', |
| |
| ' thri', |
| |
| ' FTC', |
| |
| ' Sao', |
| |
| ' Nikola', |
| |
| ' oblig', |
| |
| 'fal', |
| |
| ' grapp', |
| |
| ' Shak', |
| |
| 'STAR', |
| |
| ' Mour', |
| |
| ' SER', |
| |
| 'Bron', |
| |
| 'Stra', |
| |
| ' Kok', |
| |
| ' jan', |
| |
| ' Smy', |
| |
| ' strives', |
| |
| ' gri', |
| |
| ' chimpan', |
| |
| ' stripe', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'ri-family', 1 |
|
|
| def prog_L06F08068(input_tokens): |
| """3 token families, 6 surface forms (train F1 of the uncapped optimum: 0.037). |
| Fires at 29,228 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to identity and self-perception""" |
| |
| fire = { |
| |
| ' you', ' YOU', |
| |
| ' we', ' WE', |
| |
| ' they', ' THEY', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'you-family', 1 |
|
|
| def prog_L06F08069(input_tokens): |
| """6 token families, 9 surface forms (train F1 of the uncapped optimum: 0.069). |
| Fires at 29,336 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to academic institutions and scientific papers""" |
| |
| fire = { |
| |
| ' of', ' Of', ' OF', |
| |
| ' Awards', ' Award', |
| |
| ' Sciences', |
| |
| ' Academy', |
| |
| ' Peace', |
| |
| 'age', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'of-family', 1 |
|
|
| def prog_L06F08070(input_tokens): |
| """6 token families, 6 surface forms (train F1 of the uncapped optimum: 0.096). |
| Fires at 71,283 of 36,786,566 positions.""" |
| |
| """prev SAE description: incidents involving law enforcement and protests""" |
| |
| fire = { |
| |
| '\n', |
| |
| 'Updated', |
| |
| 'By', |
| |
| ' |', |
| |
| 'Posted', |
| |
| 'legraph', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '\\n-family', 1 |
|
|
| def prog_L06F08071(input_tokens): |
| """14 token families, 21 surface forms (train F1 of the uncapped optimum: 0.106). |
| Fires at 23,474 of 36,786,566 positions.""" |
| |
| """prev SAE description: action verbs related to processes or actions that are being taken or executed""" |
| |
| fire = { |
| |
| ' done', ' Done', 'Done', |
| |
| ' elected', 'elected', |
| |
| ' passed', ' Passed', |
| |
| ' compared', |
| |
| ' introduced', 'Introduced', ' Introduced', |
| |
| ' convicted', |
| |
| ' approved', |
| |
| ' combined', ' Combined', |
| |
| ' attacked', |
| |
| ' completed', |
| |
| ' applied', |
| |
| ' exposed', |
| |
| ' properly', |
| |
| ' confronted', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'done-family', 1 |
|
|
| def prog_L06F08072(input_tokens): |
| """25 token families, 36 surface forms (train F1 of the uncapped optimum: 0.151). |
| Fires at 81,412 of 36,786,566 positions.""" |
| |
| """prev SAE description: words and phrases related to various aspects of data and metadata""" |
| |
| fire = { |
| |
| 'a', 'ça', |
| |
| 'x', 'X', ' x', |
| |
| 'o', 'ão', |
| |
| 'z', |
| |
| 'e', |
| |
| 'en', 'EN', |
| |
| 'at', |
| |
| 'd', |
| |
| 'n', 'ón', |
| |
| 'c', |
| |
| 'k', |
| |
| 'g', |
| |
| 'st', |
| |
| 'f', |
| |
| 'j', |
| |
| 'ab', 'AB', |
| |
| '\xad', |
| |
| 'w', |
| |
| 'ig', 'IG', |
| |
| '_', |
| |
| 'iv', 'IV', |
| |
| 'te', 'TE', |
| |
| '@', |
| |
| 'iz', 'IZ', |
| |
| 'ri', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'a-family', 1 |
|
|
| def prog_L06F08073(input_tokens): |
| """19 token families, 21 surface forms (train F1 of the uncapped optimum: 0.452). |
| Fires at 6,434 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases expressing improbability or doubt""" |
| |
| fire = { |
| |
| ' likely', 'likely', |
| |
| ' unlikely', |
| |
| ' surprising', |
| |
| ' reluctant', |
| |
| ' improbable', |
| |
| 'lier', |
| |
| 'liest', |
| |
| ' presumed', ' presum', |
| |
| 'fetched', |
| |
| ' skeptical', |
| |
| ' unsur', |
| |
| 'assuming', |
| |
| 'prising', |
| |
| ' doubtful', |
| |
| ' unthinkable', |
| |
| ' dubious', |
| |
| 'ausible', |
| |
| ' pessimistic', |
| |
| 'ceivable', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'like-family', 1 |
|
|
| def prog_L06F08074(input_tokens): |
| """7 token families, 10 surface forms (train F1 of the uncapped optimum: 0.768). |
| Fires at 3,780 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases related to capital and financial investments""" |
| |
| fire = { |
| |
| ' capital', ' Capital', 'Capital', 'capital', |
| |
| ' lever', |
| |
| ' collateral', |
| |
| ' mana', |
| |
| ' capitals', |
| |
| 'phil', |
| |
| ' capitalists', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'capit-family', 1 |
|
|
| def prog_L06F08075(input_tokens): |
| """11 token families, 27 surface forms (train F1 of the uncapped optimum: 0.150). |
| Fires at 204,800 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: proper nouns, particularly geographic locations and notable individuals""" |
| |
| fire = { |
| |
| ' Trump', 'Trump', |
| |
| ' police', ' Police', 'Police', |
| |
| ' state', ' State', 'State', |
| |
| ' U', |
| |
| ' Mr', ' MR', |
| |
| ' Apple', 'Apple', |
| |
| ' Google', 'Google', |
| |
| ' Republican', ' Republicans', 'Republicans', |
| |
| ' Obama', 'Obama', |
| |
| ' Clinton', 'Clinton', ' Clintons', |
| |
| ' Bitcoin', ' bitcoin', 'Bitcoin', ' Bitcoins', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'trump-family', 1 |
|
|
| def prog_L06F08076(input_tokens): |
| """19 token families, 33 surface forms (train F1 of the uncapped optimum: 0.532). |
| Fires at 10,696 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to historical timelines or dates""" |
| |
| fire = { |
| |
| ' dating', ' dates', ' dated', ' Dates', 'dated', 'dating', ' Dating', 'dates', |
| |
| ' date', ' Date', 'Date', |
| |
| ' chron', ' Chron', 'chron', |
| |
| ' traced', ' traces', |
| |
| ' spanning', |
| |
| ' numbered', ' numbering', 'numbered', |
| |
| ' ال', |
| |
| ' valued', |
| |
| ' circa', |
| |
| ' spans', |
| |
| ' Dynasty', |
| |
| ' owes', |
| |
| '\u200e', |
| |
| ' pegged', |
| |
| ' timed', |
| |
| ' authenticity', |
| |
| 'aceous', |
| |
| ' totaling', |
| |
| 'alog', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'dat-family', 1 |
|
|
| def prog_L06F08077(input_tokens): |
| """8 token families, 20 surface forms (train F1 of the uncapped optimum: 0.087). |
| Fires at 347,135 of 36,786,566 positions. |
| Fires at 0.94% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: specific names and terms associated with individuals and organizations in a scientific and news context""" |
| |
| fire = { |
| |
| ',', ' ,', |
| |
| ' and', ' AND', 'anders', ' Anders', 'anding', |
| |
| 's', 'és', 'ís', |
| |
| ' (', |
| |
| ' or', ' OR', |
| |
| ' —', '—', |
| |
| ' -', |
| |
| 'o', ' O', ' o', 'ão', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F08078(input_tokens): |
| """14 token families, 31 surface forms (train F1 of the uncapped optimum: 0.169). |
| Fires at 91,193 of 36,786,566 positions.""" |
| |
| """prev SAE description: terms related to Jewish and Muslim communities and organizations""" |
| |
| fire = { |
| |
| ' community', ' communities', ' Community', ' Commun', ' Communities', |
| |
| ' Church', ' church', ' churches', ' Churches', |
| |
| ' leaders', ' leader', ' Leaders', |
| |
| ' groups', ' Groups', |
| |
| ' Party', |
| |
| ' rights', ' Right', 'rights', |
| |
| ' Center', |
| |
| ' marriage', ' Marriage', 'marriage', |
| |
| ' population', |
| |
| ' movement', ' Movement', |
| |
| 'ism', |
| |
| ' Association', |
| |
| ' minority', ' minorities', |
| |
| ' militants', ' militant', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'commun-family', 1 |
|
|
| def prog_L06F08079(input_tokens): |
| """29 token families, 36 surface forms (train F1 of the uncapped optimum: 0.739). |
| Fires at 7,212 of 36,786,566 positions.""" |
| |
| """prev SAE description: instances of the word "Stone." """ |
| |
| fire = { |
| |
| ' Stone', 'stone', 'Stone', |
| |
| ' Drake', |
| |
| ' Stones', 'stones', |
| |
| ' Richards', |
| |
| 'ste', |
| |
| ' artifacts', ' artifact', ' Artifact', |
| |
| 'estone', |
| |
| 'ettel', |
| |
| ' Tiger', |
| |
| ' Stark', |
| |
| ' Slate', |
| |
| ' Fields', |
| |
| ' Smithsonian', |
| |
| ' Spears', ' Spear', |
| |
| 'isman', |
| |
| 'Tri', |
| |
| ' Andrews', |
| |
| 'arton', |
| |
| 'olith', |
| |
| ' Dartmouth', |
| |
| ' Thiel', |
| |
| ' Steele', |
| |
| ' Viking', |
| |
| ' Splash', |
| |
| ' Starr', |
| |
| ' arte', |
| |
| ' Sands', |
| |
| ' Spo', |
| |
| 'ifact', 'ifacts', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'stone-family', 1 |
|
|
| def prog_L06F08080(input_tokens): |
| """25 token families, 36 surface forms (train F1 of the uncapped optimum: 0.644). |
| Fires at 6,809 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to rent and housing costs""" |
| |
| fire = { |
| |
| ' rent', ' rental', 'rent', ' rented', ' rents', ' renting', ' Rent', ' renters', |
| |
| ' landlord', ' landlords', |
| |
| ' ren', |
| |
| 'sharing', |
| |
| ' Sara', |
| |
| 'demand', |
| |
| ' zoning', |
| |
| ' lease', 'lease', |
| |
| ' lent', ' Lent', |
| |
| ' Craigslist', |
| |
| 'icted', |
| |
| ' Hert', |
| |
| ' eviction', |
| |
| ' rentals', |
| |
| 'rification', |
| |
| ' tv', |
| |
| ' vac', |
| |
| ' Jess', |
| |
| ' Sung', |
| |
| ' liv', |
| |
| ' leased', |
| |
| ' resize', |
| |
| 'ssh', ' ssh', |
| |
| 'ansom', |
| |
| ' dues', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'rent-family', 1 |
|
|
| def prog_L06F08081(input_tokens): |
| """46 token families, 61 surface forms (train F1 of the uncapped optimum: 0.701). |
| Fires at 11,426 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to courthouses and related legal contexts""" |
| |
| fire = { |
| |
| ' Cour', ' cour', 'cour', 'Cour', ' COUR', |
| |
| 'Court', ' Courts', ' COURT', |
| |
| 'oir', |
| |
| ' Hunts', |
| |
| ' courthouse', |
| |
| ' Brow', ' brow', 'Brow', |
| |
| 'BRE', ' bre', |
| |
| ' Jur', ' jur', |
| |
| 'Jes', ' Jes', |
| |
| ' arche', ' Arche', |
| |
| ' cinem', |
| |
| ' strive', ' strives', |
| |
| 'Journal', |
| |
| 'agar', |
| |
| 'aval', |
| |
| ' Cere', |
| |
| ' unab', |
| |
| ' Cove', ' cove', |
| |
| ' correspondence', |
| |
| 'brid', |
| |
| ' Hoover', |
| |
| ' Cout', |
| |
| ' Bever', |
| |
| 'thouse', |
| |
| ' tut', |
| |
| ' Jol', |
| |
| ' trou', |
| |
| 'oche', |
| |
| ' Chel', 'Chel', |
| |
| ' tresp', |
| |
| ' jog', |
| |
| ' conve', |
| |
| ' blur', |
| |
| ' Cumber', |
| |
| ' Aur', |
| |
| ' courtyard', |
| |
| ' bub', |
| |
| ' Witt', |
| |
| ' unquestion', |
| |
| ' Clair', |
| |
| ' convers', |
| |
| ' shrew', |
| |
| ' grou', |
| |
| ' ré', |
| |
| ' refres', |
| |
| ' refere', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'cour-family', 1 |
|
|
| def prog_L06F08082(input_tokens): |
| """3 token families, 3 surface forms (train F1 of the uncapped optimum: 0.075). |
| Fires at 58,948 of 36,786,566 positions.""" |
| |
| """prev SAE description: repeated phrases, particularly those introducing ideas or transitions""" |
| |
| fire = { |
| |
| ',', |
| |
| ' however', |
| |
| ' mind', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F08083(input_tokens): |
| """8 token families, 19 surface forms (train F1 of the uncapped optimum: 0.103). |
| Fires at 114,390 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to violence and hate crimes""" |
| |
| fire = { |
| |
| '"', |
| |
| ' against', ' Against', 'against', |
| |
| "'", |
| |
| 'ist', 'ists', 'istic', |
| |
| ' violence', ' Violence', 'violence', |
| |
| ' attack', ' attacks', ' Attacks', |
| |
| ' crime', 'crime', ' Crime', |
| |
| 'ism', 'ISM', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '"-family', 1 |
|
|
| def prog_L06F08084(input_tokens): |
| """43 token families, 61 surface forms (train F1 of the uncapped optimum: 0.617). |
| Fires at 11,642 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to "clusters," indicating groupings in various contexts""" |
| |
| fire = { |
| |
| ' cluster', ' clusters', ' Cluster', ' clust', |
| |
| ' hub', ' hubs', |
| |
| ' clan', ' Clan', ' clans', |
| |
| 'uster', 'usters', |
| |
| ' clin', ' Clinical', ' Clin', |
| |
| ' Rift', |
| |
| 'uddle', |
| |
| ' swarm', ' Swarm', |
| |
| ' grouped', ' grouping', 'Group', 'groups', |
| |
| ' Guatemala', |
| |
| ' CU', 'CU', |
| |
| ' Cisco', |
| |
| ' gangs', |
| |
| ' constellation', |
| |
| ' cores', |
| |
| ' baff', |
| |
| ' hots', 'hots', |
| |
| ' Kosovo', |
| |
| 'uddled', |
| |
| ' Cra', |
| |
| ' Yun', |
| |
| ' Clash', |
| |
| ' aura', ' Aura', |
| |
| ' Clover', ' cloves', |
| |
| ' Starr', |
| |
| ' HA', |
| |
| ' Conc', |
| |
| ' clut', |
| |
| ' Kara', |
| |
| ' pods', |
| |
| ' densely', |
| |
| ' Killer', |
| |
| ' clutch', |
| |
| ' clinch', |
| |
| ' Chomsky', |
| |
| ' Lena', |
| |
| ' Nebula', |
| |
| ' oddly', |
| |
| 'lust', 'luster', |
| |
| ' cabal', |
| |
| 'ellation', |
| |
| ' stink', |
| |
| ' cursor', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'clust-family', 1 |
|
|
| def prog_L06F08085(input_tokens): |
| """19 token families, 23 surface forms (train F1 of the uncapped optimum: 0.731). |
| Fires at 10,786 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to effort and the value associated with it""" |
| |
| fire = { |
| |
| ' effort', ' efforts', |
| |
| ' attempt', ' Attempt', |
| |
| ' strive', ' strives', |
| |
| ' endeavor', |
| |
| ' foray', |
| |
| ' Eff', 'eff', |
| |
| ' concerted', |
| |
| ' striving', |
| |
| 'ermott', |
| |
| ' uncom', |
| |
| ' exert', |
| |
| 'Labor', |
| |
| 'rahim', |
| |
| 'USH', |
| |
| ' endeavour', |
| |
| ' comput', |
| |
| 'illian', |
| |
| 'ERSON', |
| |
| 'Push', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'effort-family', 1 |
|
|
| def prog_L06F08086(input_tokens): |
| """42 token families, 45 surface forms (train F1 of the uncapped optimum: 0.530). |
| Fires at 6,121 of 36,786,566 positions.""" |
| |
| """prev SAE description: the concept of being at a midpoint or halfway mark in various contexts""" |
| |
| fire = { |
| |
| ' halfway', |
| |
| ' joked', ' joking', |
| |
| ' trilogy', |
| |
| ' safely', |
| |
| ' midway', |
| |
| ' halftime', |
| |
| ' dangerously', |
| |
| ' pronouns', ' pronoun', |
| |
| 'loop', |
| |
| ' cru', |
| |
| ' anal', |
| |
| ' nearing', |
| |
| ' prol', ' Prol', |
| |
| ' Paleo', |
| |
| 'Fin', |
| |
| ' fart', |
| |
| ' cruising', |
| |
| ' infinitely', |
| |
| ' epis', |
| |
| 'ultimate', |
| |
| ' Ways', |
| |
| ' parap', |
| |
| ' Founders', |
| |
| ' reckoned', |
| |
| ' borderline', |
| |
| ' encamp', |
| |
| ' metast', |
| |
| ' controvers', |
| |
| 'binary', |
| |
| ' Spart', |
| |
| ' Doomsday', |
| |
| 'omsday', |
| |
| ' traveller', |
| |
| ' hypot', |
| |
| ' LEG', |
| |
| ' thirds', |
| |
| ' Donation', |
| |
| 'wic', |
| |
| 'comed', |
| |
| ' Oath', |
| |
| ' Apost', |
| |
| 'ollo', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'halfway-family', 1 |
|
|
| def prog_L06F08087(input_tokens): |
| """36 token families, 49 surface forms (train F1 of the uncapped optimum: 0.595). |
| Fires at 11,596 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to art exhibitions""" |
| |
| fire = { |
| |
| ' display', ' displayed', ' displays', ' displaying', |
| |
| ' exhibition', ' Exhibition', |
| |
| ' exhibit', ' exhibited', ' exhibits', ' exhibiting', ' Exhibit', |
| |
| ' showcase', |
| |
| ' showcasing', ' showcases', ' showcased', |
| |
| ' Expo', |
| |
| ' attraction', |
| |
| ' artifacts', ' artifact', |
| |
| ' attractions', |
| |
| 'hib', |
| |
| ' sculpture', |
| |
| ' Tate', |
| |
| ' housed', |
| |
| ' Pavilion', |
| |
| ' tasting', |
| |
| ' curator', |
| |
| ' retrospective', |
| |
| ' exhibitions', |
| |
| ' redevelopment', |
| |
| ' concession', |
| |
| 'hibit', 'hibited', |
| |
| ' catalogue', |
| |
| ' aur', |
| |
| 'posium', |
| |
| ' concessions', |
| |
| ' pav', |
| |
| ' renovations', |
| |
| ' Annex', ' annex', |
| |
| ' facade', |
| |
| ' Madame', |
| |
| ' exhib', |
| |
| 'iferation', |
| |
| ' amph', |
| |
| 'useum', |
| |
| 'unda', |
| |
| ' exposition', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'display-family', 1 |
|
|
| def prog_L06F08088(input_tokens): |
| """37 token families, 59 surface forms (train F1 of the uncapped optimum: 0.690). |
| Fires at 8,758 of 36,786,566 positions.""" |
| |
| """prev SAE description: variations and instances of the word "bow." """ |
| |
| fire = { |
| |
| ' bow', ' Bow', 'bow', 'Bow', ' bowed', 'bows', ' bows', ' Bows', |
| |
| ' bowl', ' bowling', ' Bowling', |
| |
| ' beams', ' beam', ' Beam', |
| |
| ' Barker', ' bark', ' barking', |
| |
| ' Bowie', |
| |
| ' bats', |
| |
| ' bend', ' bending', ' bends', |
| |
| ' arrow', ' arrows', |
| |
| ' Brave', 'Brave', |
| |
| ' WB', 'WB', |
| |
| ' BST', |
| |
| ' Braves', |
| |
| ' kne', |
| |
| ' Hodg', |
| |
| ' Bowen', |
| |
| ' Dexter', |
| |
| ' Pear', ' pear', 'Pear', |
| |
| ' Bowman', |
| |
| ' bowel', |
| |
| ' Alas', ' alas', |
| |
| 'aughty', |
| |
| 'bb', |
| |
| ' violin', |
| |
| ' Dart', |
| |
| 'embed', |
| |
| ' Archer', |
| |
| ' hem', |
| |
| ' Bravo', |
| |
| '........', ' ........', |
| |
| ' bellig', |
| |
| ' bladder', |
| |
| ' Brah', |
| |
| ' prow', |
| |
| ' beck', |
| |
| ' seam', |
| |
| ' adj', |
| |
| ' Blossom', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'bow-family', 1 |
|
|
| def prog_L06F08089(input_tokens): |
| """25 token families, 29 surface forms (train F1 of the uncapped optimum: 0.289). |
| Fires at 31,456 of 36,786,566 positions.""" |
| |
| """prev SAE description: comparisons emphasizing improvement or superiority""" |
| |
| fire = { |
| |
| ' better', ' Better', 'better', |
| |
| ' similar', ' similarly', |
| |
| ' potential', |
| |
| ' larger', |
| |
| ' higher', 'higher', |
| |
| ' bigger', |
| |
| ' additional', |
| |
| ' extra', |
| |
| ' particular', |
| |
| ' lower', |
| |
| ' greater', |
| |
| ' separate', |
| |
| ' fresh', |
| |
| ' closer', |
| |
| ' nearby', |
| |
| ' smaller', |
| |
| ' mere', |
| |
| ' wider', |
| |
| ' broader', |
| |
| ' stronger', |
| |
| ' deeper', |
| |
| ' cheaper', |
| |
| ' lesser', |
| |
| ' renewed', |
| |
| ' shorter', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'bett-family', 1 |
|
|
| def prog_L06F08090(input_tokens): |
| """6 token families, 7 surface forms (train F1 of the uncapped optimum: 0.274). |
| Fires at 14,889 of 36,786,566 positions.""" |
| |
| """prev SAE description: phrases emphasizing comparison and degree, particularly relating to "more than anything" or "more than any other." """ |
| |
| fire = { |
| |
| ' any', ' ANY', |
| |
| ' anyone', |
| |
| ' anything', |
| |
| ' else', |
| |
| ' anywhere', |
| |
| ' anybody', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'any-family', 1 |
|
|
| def prog_L06F08091(input_tokens): |
| """23 token families, 60 surface forms (train F1 of the uncapped optimum: 0.326). |
| Fires at 62,253 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to home buying and real estate""" |
| |
| fire = { |
| |
| ' home', 'home', |
| |
| ' housing', ' houses', ' Housing', 'houses', 'housing', |
| |
| ' house', |
| |
| ' homes', ' Homes', |
| |
| ' property', ' Property', 'Property', |
| |
| ' living', ' lived', ' liv', |
| |
| ' apartment', ' apartments', ' Apart', |
| |
| ' estate', ' Estate', 'estate', |
| |
| ' units', 'unit', 'units', |
| |
| ' sold', |
| |
| ' rental', ' rent', ' renting', ' rented', ' rents', ' Rent', ' renters', |
| |
| ' residential', ' Residential', 'resident', |
| |
| ' properties', ' Properties', |
| |
| ' sale', |
| |
| ' sell', |
| |
| ' buildings', ' builder', 'builders', |
| |
| ' flat', ' flats', |
| |
| ' buyers', 'buy', ' buyer', |
| |
| ' homeowners', ' homeowner', |
| |
| ' affordable', ' affordability', |
| |
| ' tower', ' towers', ' Towers', |
| |
| 'bedroom', ' bedroom', ' bedrooms', |
| |
| ' condo', ' condos', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'home-family', 1 |
|
|
| def prog_L06F08092(input_tokens): |
| """6 token families, 10 surface forms (train F1 of the uncapped optimum: 0.064). |
| Fires at 47,696 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to extending agreements or support in different contexts""" |
| |
| fire = { |
| |
| ' its', |
| |
| ' extended', ' extend', ' extending', ' extends', |
| |
| ' contract', ' contracts', |
| |
| 'able', |
| |
| ' ban', |
| |
| 'term', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'its-family', 1 |
|
|
| def prog_L06F08093(input_tokens): |
| """48 token families, 59 surface forms (train F1 of the uncapped optimum: 0.688). |
| Fires at 13,191 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to specific years in the 1800s""" |
| |
| fire = { |
| |
| ' 1950', '1950', |
| |
| ' 900', '900', |
| |
| ' 1930', |
| |
| ' 1800', '1800', |
| |
| ' 1900', '1900', |
| |
| ' 1500', '1500', |
| |
| ' 2050', |
| |
| ' Maur', |
| |
| ' ACT', 'Act', |
| |
| ' 1940', |
| |
| ' 1600', '1600', |
| |
| ' 1200', '1200', |
| |
| ' noon', |
| |
| ' 1920', '1920', |
| |
| 'alth', 'ALTH', |
| |
| ' Mons', |
| |
| ' Maxwell', |
| |
| ' 850', '850', |
| |
| ' 1910', |
| |
| ' 1890', |
| |
| 'mouth', |
| |
| ' populous', |
| |
| ' abolition', |
| |
| 'mount', |
| |
| ' 1700', |
| |
| ' 1880', |
| |
| ' unders', |
| |
| ' autobiography', |
| |
| ' Hust', |
| |
| ' midday', |
| |
| ' nineteenth', |
| |
| 'ampunk', |
| |
| ' measles', |
| |
| ' Martial', |
| |
| ' methamphetamine', |
| |
| ' skirm', |
| |
| ' deforestation', |
| |
| ' 1898', |
| |
| ' 1860', |
| |
| ' 1300', |
| |
| ' 1870', |
| |
| ' 1850', |
| |
| ' 1905', |
| |
| ' 1904', |
| |
| ' landfall', |
| |
| ' 1899', |
| |
| ' 1903', |
| |
| ' 1830', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return '1950-family', 1 |
|
|
| def prog_L06F08094(input_tokens): |
| """2 token families, 3 surface forms (train F1 of the uncapped optimum: 0.176). |
| Fires at 156,772 of 36,786,566 positions. |
| Fires at 0.43% of positions, and |
| token rules do not serve features this dense: median test F1 0.086 |
| across the 37,173 features above 0.39% of positions, against 0.763 for |
| the 76,283 below 0.016%. Read this fire set as the best available |
| approximation, not as what the feature does.""" |
| |
| """prev SAE description: instances of commas in a text""" |
| |
| fire = { |
| |
| ',', |
| |
| '",', ' ",', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return ',-family', 1 |
|
|
| def prog_L06F08095(input_tokens): |
| """14 token families, 20 surface forms (train F1 of the uncapped optimum: 0.650). |
| Fires at 7,246 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of leadership positions or titles""" |
| |
| fire = { |
| |
| ' chair', ' Chair', ' chairs', 'chair', ' chaired', 'Chair', 'chairs', |
| |
| ' chairman', |
| |
| ' congress', |
| |
| ' conven', |
| |
| ' organ', |
| |
| 'Joined', |
| |
| ' stair', |
| |
| 'found', |
| |
| ' cha', |
| |
| ' spoon', |
| |
| ' emer', |
| |
| ' Gentle', |
| |
| 'spons', |
| |
| 'cand', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'chair-family', 1 |
|
|
| def prog_L06F08096(input_tokens): |
| """12 token families, 22 surface forms (train F1 of the uncapped optimum: 0.167). |
| Fires at 96,695 of 36,786,566 positions.""" |
| |
| """prev SAE description: concepts related to trust, accountability, and emotional states in various contexts""" |
| |
| fire = { |
| |
| ' support', |
| |
| ' development', ' Development', |
| |
| ' security', |
| |
| ' analysis', ' Analysis', |
| |
| ' power', |
| |
| 'ability', ' ability', ' abilities', |
| |
| ' technology', ' Technology', |
| |
| ' culture', ' Culture', 'culture', |
| |
| ' attention', |
| |
| ' violence', 'violence', |
| |
| ' abuse', ' Abuse', |
| |
| ' safety', 'safety', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'support-family', 1 |
|
|
| def prog_L06F08097(input_tokens): |
| """26 token families, 33 surface forms (train F1 of the uncapped optimum: 0.853). |
| Fires at 15,992 of 36,786,566 positions.""" |
| |
| """prev SAE description: the word "am" in various contexts""" |
| |
| fire = { |
| |
| 'am', 'AM', 'Am', |
| |
| 'ams', 'AMS', |
| |
| ' noon', |
| |
| 'Sam', 'sam', |
| |
| ' Nam', |
| |
| 'amin', |
| |
| 'amy', |
| |
| ' Alam', 'alam', |
| |
| ' Mam', |
| |
| ' Cham', 'cham', |
| |
| ' tram', |
| |
| 'amon', |
| |
| ' Pam', |
| |
| 'cam', |
| |
| ' Zam', |
| |
| ' Pamela', |
| |
| 'amaz', |
| |
| 'aram', |
| |
| 'amel', |
| |
| 'andel', |
| |
| 'umar', |
| |
| ' Beam', 'beam', |
| |
| ' Bam', |
| |
| 'ndum', |
| |
| ' Salam', |
| |
| 'amus', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'am-family', 1 |
|
|
| def prog_L06F08098(input_tokens): |
| """19 token families, 32 surface forms (train F1 of the uncapped optimum: 0.703). |
| Fires at 9,216 of 36,786,566 positions.""" |
| |
| """prev SAE description: mentions of the product "Windows" and its various versions""" |
| |
| fire = { |
| |
| ' win', ' Win', 'win', ' wins', 'Win', ' Wins', 'WIN', ' WIN', |
| |
| ' Won', 'won', |
| |
| ' KO', 'KO', |
| |
| ' Winston', |
| |
| ' Luck', 'Luck', |
| |
| ' Winn', ' Winning', 'Winner', |
| |
| ' Earn', 'Earn', |
| |
| ' Wen', |
| |
| ' Vin', |
| |
| ' lyn', |
| |
| ' Oprah', |
| |
| ' Mae', |
| |
| ' Kop', |
| |
| ' Isa', |
| |
| 'Joy', |
| |
| ' POW', |
| |
| 'GRE', |
| |
| ' dow', |
| |
| 'yna', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'win-family', 1 |
|
|
| def prog_L06F08099(input_tokens): |
| """15 token families, 25 surface forms (train F1 of the uncapped optimum: 0.111). |
| Fires at 54,216 of 36,786,566 positions.""" |
| |
| """prev SAE description: references to ease of use or usability in various contexts""" |
| |
| fire = { |
| |
| ' use', 'use', ' Use', ' USE', |
| |
| ' get', |
| |
| ' play', 'play', |
| |
| ' keep', |
| |
| ' find', 'find', |
| |
| ' share', |
| |
| ' read', |
| |
| ' build', ' Build', |
| |
| ' understand', |
| |
| ' watch', |
| |
| ' learn', |
| |
| ' buy', |
| |
| ' write', ' Write', |
| |
| ' maintain', ' maintaining', |
| |
| ' install', ' Install', 'install', |
| } |
| token = input_tokens[-1] |
| if token not in fire: |
| return "no match", 0 |
| return 'use-family', 1 |
|
|
