III. Lo Scrittoio
ArticlesNotes from the scriptorium
. : - = + # % @ % # + = - : .
· 13 min
Hallucinations in language models
How a language model can give you a good explanation of something that never happened, and what you can do to spot it.
/ / /
| / / / / / / / /
| | / | / // / / | | | | | | /
/ // | / / // | | /
/ / | / | | ||
/ / / / / | |
| | / / | | /
/ / / / | / / /
| / / | | / | |
/ | / / | / /
/ | / | /
| / // / / / / / | / | /
/ / / / // | / / /
/ | | / | / | |
| | / / / / /
/ | |
| / / / /
/ /
| |/ / / | / / |
| / | /| / | | /
/ // / / / / / /
/ / / / / / / | /
/ / /
/ | / / / | |
/ | / / / / / /
|/ / / | / / / | | |
/ / / | / /| | // / /
/ | / / / /
| / / | / / / | | / /
| | / / / / / |
| / | / / | / / /
/ / | / / | | / / /
/ / / / / / / |
| / / /
| | /| / / / / | | / /
/ | / / / | / / / /
/ / // / / / // | / // |
/ / | | | | / /
/ | / / | / /
/ | | | / | /
/ / / | / / / / / / | |
| / / / | / / | |
/ | / / / /
/ / / |
/ // || / /
// / | / //
|/ / / / |/ | /
| | | | / | / | / / / | / |
// / / / / | |/ | / /
/ / | / / / | / / |
/ / / / | |
| / | / /
/ / / /
| / | | | |
/ / / / /
/ / / / | /
/ / | /
| / /
| /| |
/ / | |
| // / / / |
/ / / / / /
/ / / | /
//
/ | / / /
| / / / / | / / / |
| / | / /
/ / /
/ / / |
/ / / / / |
/ / / /
/ /
/ / / / | |
| / |
/ / | /
| / / / | | / / /
/ / / /
/ | // /
/ / / / / | /
/ / / /
// / | / | / ||
/ / / / | | | / / / / /
/ / //
| / / / /
/ / / / / / / |
/ |
| / / / / / /
/ / / / / / / | /
| |/ / / | |
//
/ | /
/ / / / / /
| | |
/ / | | / / / | |
| | | | |
/ / / / / /
/ | // / / /
/ /
/ / / / / / |
| | | / /
| | / /
/ / / / / /
| |
================================================================================================================================================================================
| |
| |
| |
| |
| P(neste ord | ledetrådene) |
| |
| |
| |
| |
| |
| |
| |
| ======== |
| |
| |
| |
| ===== |
| |
| |
| |
| == |
| |
| |
| |
| |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : |
| : +-----+ +---------+ +---+ +--------+ +----+ +-----+ +-----+ |
| : |
| : +-----+ +---------+ +---+ +--------+ +----+ +-----+ +-----+ |
| : |
| : +-----+ +-----+ |
| : |
| : +-----+ +-----+ |
| : |
| |
| |
| |
| |
| |
| |
| |
================================================================================================================================================================================
^
@ jakken
@% @ %@
% @ %
@ @ %
#@ # @ @ ## regnjakken
#@ @ @ % @#
# # @ % @
% @# @ %% #
@@% @ @ @ #%@ hatten
@# @ @ # %@
% @ @
@
@@@@@@@@@ @@@@@@@@@ ######### ######### #
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| | Det | | begynte | | å | | regne, | | så | | han | | tok |
|
|
|
| | med | | seg |
|
|
. o
o. ..
...
+ +
paraplyen #####################################
/// =--=--=-=-= =--=--=
///xxx### -==-=-=-=- =-#---=---
/x#%#%%%%%##x #==#-#-== =#==-==#=
//x# xxx###x#x#% -#-#-#-= %#=-=-== xx\
x/x##xxx# #####%%@# #=#==## ======= xxxx\\
xx/x#xx#%xxx# @@@@@%## -#==== #==#%# \xxxxxx\ \\x\
/x#x##xx%#x#%##x @@%#%@###=== %#####\x#x\\ \xxx\\x
|%##% #x#%x#%#x%%#x ###@%%===# @####\#\#x \x\\x#\\x\\
|%|%%#%@#@@%@ %#x%%x#%%x #@%@%## @#####x x%#%##x\xx
#|%##%@#@@#@@#@%%@%% x%#x%%#@%@== %%xxx#x%#x## ||||||
|%%##@%#@@#@@%@@#@@%@%@@%@%%x%### #x%%%#|||#|||||||||
+ - - - - - - - +
| paraplyen ? |
+ - - - - - - - +
+ +
--:----:-: -----:-::-
--=- \..
\::\:..:.
:\\\:::\:.::
\\::\\:. .::
:\\ \\\::\.:.
\\\.\:::.
\\\:\.\.\
x: : ..
:|:|.|:::.:..
||||:|.::.:: :
|||||:||::|..:
· 12 min
How language models work
An explanation without the math: how a language model guesses its way forward one word at a time, what attention really is, and why it can sound convincing even when it is wrong.
/skills/docx/SKILL.md /skills/stilguide/SKILL.md /skills/design/SKILL.md /skills/stilguide/SKILL.md /skills/valider/SKILL.md /skills/pdf/SKILL.md /skills/budsje
/skills/xl / je pi s/d x/SKILL.md /skills/api-docs/SKILL.md /skills/api-docs/SKILL.md /skills/xlsx
/skills/me i n L s il e i .md /skills/refe KILL.md
/skills/do / L k s /SKILL.md /skills/xl s /SK L.md /skills/meeting-notes/SKILL.md /s ills/stilguide/SKILL.md /skills/desi
/skills/st g /referat kil /meeting-note ILL.md /
/skills/de g K / p tx m k id . s l omm /SKILL.md /skills/meeting-notes/SKILL.md /skills/docx/SKILL.md /skills/desi
/skills/pp / /S ng ls/ dsjett/SKILL. skills/des
/skills/co i L / s il K s/ / ill docx/SKILL.md /skills/docx/SKILL.md /skills/valider/SKILL.md /skills/meetin
/skills/budsjett/SKILL.md /skills/xlsx/SKILL.md /skills/pptx/SKILL.md /skills/xlsx/SKILL.md /skills/xlsx/SKILL.md /skills/valider/SKILL.md /skills/valider/SKILL
: : :|. :: :..:.. :: .| |: ::. .::. |: .:: :.::.. :|
:. .:: :... : | : .:. .:. .| ::. .::
| : .::. .:...| ::. ::.. .|. .| . .::. | |...:
.|. .| | .::. . | |. .:. ..:. : ...: . ::. :|
..:: : |:. |: :. ::: :| :..::. --------- |. ::: ..:..:
:. :: ...: . ::. :..:.. .:: .: :..: :: .::
: .| |. ::: ..:.. :: .:: : ::.. : --------- ::. .:...| |
---------
..:: : |:. :: :. --------- |: ::: ..:..:
:. ::. .| :. :. : .::.
: .| |: ::. ..|.. .::. .|: .| |
:.. .:. | : .:.. .|: ..: : ::
| |..:.. .|. .| : |:. :: :..:
|. .|. ..:: : . ..: .| |: .:
::: :| :. :: :. |: ::. ..|..:
|. .::. | |: .|: ::.: :| |. ::. ..:
|: . .:: : :: ::: .: :.. |: .:
::: :| :. ::. . | |: ::: ..:..: .:| : ::.. .|..:|
. :.. . : .: |: ..:. : :: .:: : . :|. .| | .:.
: .:: : ::.. :|. | .::. .::..| |. .::. | |: .|:
:..| |. .:. .:. |. .|: . .: : ::. |: .:. ..:: : ::
:. .::. : |:..|: :: : ::. :: :..:. ::: :| |. ::. .::
| .:: :.::.. :|. |..::. .|:..| |. .::. : |:..|:
: .| ..:. |: :: : ::. : .: : .| :
:.: ::. :| :..|:. :: :| :. ::: :..:. ::: :| |. ::. .:|
: :..:.. .:: .: ::. ..: : |: .:: :... :: .:: : .:
:: .:: ..:..: :| : ::.. .|. .| .::. .::..| |. .|
..:. . :| .:: : .|. .| | .::. .| |. .:. ..:. :
| .::. .::..| |. ..:: : |:..|: . : ::. :| :..|:.
================================================================================================================================================================================
| |
| |
| |
| |
| --------------------------------------------------------------------------------------- |
| |
| |
| |
| |
| ||. ::| |.| :|: : .|| |:: |
| . . |:. .:| :: :|: |
| ::| ||: :: :|: ||: :|| ::: |
| |: ::o :|: |:: o|| ~~~~~~~~~ | o ||. |
| | |:. ::| ||. ::| : :|| | navn + beskrivelse |
| .:| ||. | :|: |.: ~~~~~~~~~ |:: .|| |
| |.: :|: .|| ||. ||: ::: |
| ~~~~~~~~~ | |
| | |
| ~~~~~~~~~ | |
| | ||. ::| .|| | | |
| :| |.: ~~~~~~~~~ :| || | |
| |.| .|| ||: ::: v |
| |:. o .:| ::|o |:. |--------------------------------------------------------| |
| .:| ||: :: ||. ::| | | |
| |:. .|| ||. :|: |.: | | |
| |. :.| .:| .| .|| | | hele SKILL.md | |
| | | |
| |--------------------------------------------------------| |
| | |
| :|| |:. : | :.| :|: | |
| |:. |:. .:| : :| .: | |
| |. :.| :|: |.: .|| |: .|| ||: | |
| : :|: o |.: || |:o .|| |:. o:| :| | |
| .|| ||: | ||: ::: :|| |:. :|| v |
| ::: :|| .:| ||. . .:| |:. |--------------------------------------------------------| |
| :|| |:: : ||. ::| :.| :|: | | |
| | | |
| | scripts/ references/ assets/ | |
| | | |
| .|| ||: | ||: ::: :|| |:: :|| |--------------------------------------------------------| |
| :: || ||. :| :.| |
| ||. ::| |.| :| :.| :|: |
| :.| o|. :. :|: o |.: |:. o .:| |
| |:: .|| |:: .:| ||: : :|: |:: |
| | |:. .:| ||: ::| :|: ::: :|| | |
| ||: |:: || |:. ::| ||. ::| :| |
| |
| |
| |
| |
| |
| |
| |
================================================================================================================================================================================
@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@
%################################################################################
@# ##
%# ======================= ======================= ======================= ##
@# # # | | | | ##
%# # meeting-notes # | pptx | | docx | ##
@# # # | | ##
%# # # | | | | ##
@# # # | | | | ##
%# # # | | | | ##
@# # # | | | | ##
%# ####################### ####################| |################# ##
@# | | ##
%# ==#===#===#===#===#===# | |==#===#===#===#== ##
@# # # | | | ##
%# # xlsx # | | pdf | ##
@# # # | ##
%# # # | ##
@# # # meeting-notes/ | ##
%# # # | ## LAG 2 ved utlosning
@# # # | ##
%# =-=-=-=-=-=-=-=-=-=-=-= =-=-=-=-=-=-=-=-=-= ##
@# ##
%# ======================= ====== =================== ##
@# # # | | ##
%# # commit # | stilguide | | budsjett | ##
@# # # | | | | ##
%# # # | | | | ##
@# # # | | | | ##
%# # # | | | | ##
@# # # | | | | ##
%# ####################### ####################### ####################### ## LAG 3 ved behov
@# ##
%# #=#=#=#=#=#=#=#=#=#=#=# =#=#=#=#=#=#=#=#=#=#=#= #=#=#=#=#=#=#=#=#=#=#=# ##
@# # # | | | | ##
%# # api-docs # | referat | | valider | ##
@# # # | | | | ##
%# # # | | | | ##
@# # # | | | | ##
%# # # | | | | ##
@# # # | | | | ##
%# =-=-=-=-=-=-=-=-=-=-=-= -=-=-=-=-=-=-=-=-=-=-=- =-=-=-=-=-=-=-=-=-=-=-= ##
@# ##
%################################################################################
#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#
+ +
|--------------------------------------------------------|
--- name | |
| LAG 1 alltid i kontekst |
| |
| |
|--------------------------------------------------------|
@%@%@%@%@%@%@%@%@%@%@%@
#|||||||:|||:|||:|||||%
%:| |||||%
#|||||||||||:|||:|||||%
%||SKILL.md|||||||||||%
#|||||||:|||:|||:|||||%
%:||||||||||||||||||||%
#|||||||||||:|||:|||||%
%#%#%#%#%#%#%#%#%#%#%#%
+ +
· 5 min
How Claude Skills work
Skills are self-contained instruction sets Claude loads on demand. What they’re made of, how they load in layers, and why they turn Claude into a specialist in your field.
53,16768,1900,972,120674,19112,2541,1705,86818,19177,7430,158429,11,123126,93775,2085,6323,10726,154714,56976,2541,1796,12114,613,6964,59675,1305,1452,1476,7086,1037,2541,3
2662,24491,50275,1452,1476,650,3944,411,1745 ,59576,264,13,3983,991,638,3148,2807,6964,155515,59079,11,55719,2085,64714,11,1111,334,163436,395,44883,87936,599,411,6212,3825
11,1796,101740,36784,14735,322, 363,2085,20341,12,2243,8068,5806
22823,41026,1401,11,26134,29 1,18173,11,49921,1796,150794,
1347,1475,91392,15570,13,1 7 4,3453,2674,49138,102300,11,
27603,19177,158502,5770,13 1705,86818,19177,7430,15842
1452,1476,7086,1037,2541,345 ,3944,411,17458,59576,264,13,3
44883,87936,599,411,6212,38258, 3 322,7466,1072,13965,19474,11,163
154961,11,178001,805,76807,11,22823, 6 34,292,1401,2085,49764,145982,85620,3
575,7967,1900,334,3535,5172,1347,1475,91392,15570,13,1754,1066,48124,169082,259,16023,17764,3453,2674,49138,102300,11,1014,1291,1401,1452,117302,2085,2750,259,11,2085,1625,
================================================================================================================================================================================
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| "strawberry" -> gpt-4o -> 3 tokens |
| --- =- - |
| --===== x\#\\\==--- |
| --- :\\x \xx | x\\ \\\\ =- |
| \ \: | \:\ :: |
| | |
| . |
| |
| |
| ' |
| 302 1618 19772 |
| |
| . |
| |
| |
| ' |
| how many r's? |
| |
| . |
| |
| |
| ' |
| the model never sees the letters, |
| only the token ids. |
| . |
| |
| |
| ' |
| [16876] -> [ -0.332, -0.114, 0.640, 0.045, 0.854, ... 768 ] |
| |
| . |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
================================================================================================================================================================================
/
|
|
|
|
##% | %##
-=-==#%##
. . . . +-----+
| raw |
+-----+
' ' ' ' '
. . . .
' ' ' ' '
. . . . s t a w b e y
' ' ' ' '
. . . .
' ' '
. .
' '
.
+ +
@% @% |# \\
# #%@% @@@% #### \
%%#@@%%@@@@@# |%%#xx\\
### @%#%@ %#@@% #%#|x xx\ +----+ +-------+
xx#@@#%@@@%#%%@#%#@#|x##x\ | st | | berry |
x%%%#%@@%#####@%##%#|x##\\x +----+ +-------+
#% ##@@@ ##%## @#%#@ |x#x\ x
x#x##@@%#%%@%@##%%##%#|x##\\x
xx%@@###@@@@@@@#%#%|%#||##x\x
x%%### @@@%% @%##% |##|| #x\\
x####%@@%%###%%@#%%%|%##|xx#\\
//##@%%########%%#%#|#%#|\xxx\
x#%% ##%%@ @@%#% #%%|# %#x\\ x
xxxx##@@@%@%@@#|@|%#||####x\\\
x#%%@%######@@#%#%#|||||#x#\\
####x# %#%## @#%#% ||||| \\xx r r r
/x##%%@@@@@%##@#%%#|||||xxx\ ^ ^ ^
##%#######@%#%#%%######xx\x
xx ####% ##@%# @###% |\\\\
x#%%%@%@@%x@|@###%#||#xx\
xxxx#x###%%x@%#|#%||#\\\x
xx# %%%## %#||| #|#|x \
######@%#%###|###|#\x\
/xx###x#%#%@%##|#|\x\
x### @%#%# ###|| #x\
/xx#x#%x##|%|||#x\
x##%%#%##%###||\\
/xxx #%%|| ||\\
x%%###x%##||\x
/xx%#%|#|#|\
x%#% %|#||
/x%%##||x
x%%###|x
x### |
+ +
:\
x\::
:\ \ :
x\:\...
\:\...
\\:: :
\\\.:..:
\\\:\...
x \.\ :
x\:\.:..:
\\:\.: :
\\\: .:..
x:\:....:
\\:\:..:
\ \....
\\\:::...
\\.\: .
\:\:. :
\\\:...
\\:..: :
:\:. :..
\::.::
x\:...:
\\ . ::
:\:.
\\...
x :.::
\\:.
:\: .
\...
x\.
::.
\..
· 11 min
How tokens work in language models
Why can’t ChatGPT count the letters in “strawberry”? The answer is tokens: the pieces of text the model actually reads, never the letters themselves.
p=0.742 z=1.14 top_p=0. p=0.6 p=exp(z/T)/Σexp(z/T) T=0.6 p=0.182 p=0.16 top_p=0.6 p=0.002 T=0.6 T=0.6 T=0.6 T=0.6 T=
T=0.6 p=0.332 z=0.62 T= . = (z/T)/Σexp(z/T) top_p=0.6 T=1.0 top_p=0.6 T=0.6 p=0.246 T=1.0 top_p=0.6 p=0.748 top_p=0.6 T=1.0
T=0.6 Σp≥0.6 T=1.0 T=0. 0. 0.6 T=1.0 p=exp(z/T)/Σexp(z/T) p=0.638 T=0.6 T=0.6 p=exp(z/T)/Σexp(z/T) p=exp(z/T)/Σexp(z/T) top_p=0
T=0.6 top_p=0.6 T=1.0 T . top_p=0.6 p=0.782 p=0 z=1.6 6 p_p=0.6 z=-1.41 z=-1.24 top_p=0.6 z=0.28 Σp≥0.6 T=1
p=0.346 T=0.6 T=1.0 Σp≥ . to p=0.6 Σp≥0.6 T=1.0 z=0.74 T=0.6 z=-2.04 T=0.6 top_p=0.6 T=0.6 top_p=0.6 top_p=0.6 p=exp(z/T)/Σ
T=1.0 T=1.0 z=2.14 p=ex 0 Σp≥0.6 T=0.6 z=-0.49 T=0.6 T=0.6 T=1.0 top_p=0.6 Σp≥0.6 Σp≥0.6 T=0.6 z=-0.04 T=0.6 T=1.0 T
T=1.0 z=-2.14 T=1.0 top 6 T z/ /Σexp(z/T) T=0.6 top_p=0.6 z=0.65 p=exp(z/T)/Σexp(z/T) T=1.0 T=1.0 T=1.0 Σp≥0.6 Σp≥0.6 z=-0.27
p=0.015 z=1.85 top_p=0.6 = 65 z=-2.6 z=1.67 T=0.6 top_p=0.6 top_p=0.6 p=exp(z/T)/Σexp(z/T) p=exp(z/T)/Σexp(z/T) p=exp(z/T)/Σexp
p=exp(z/T)/Σexp(z/T) p=ex .87 z=2.37 Σp≥0.6 p=exp(z/T)/Σexp(z/T) p=exp(z/T)/Σexp(z/T) T=1.0 z=-2.65 p=exp(z/T)/Σexp(z/T) Σp≥
.:. .:..::.:|: ::. ::
.:. ::. ::..|...:...: .
.:. .|..:|.:|: ::. ::.
.:. ::. ::..|...: .:
.:. .:..::.: ::
.:. ::. :: .
.:. .|..:| .
.:. ::. ::
.:. .:..::
.:. ::. :: .
.:. .|..:|.: ::.
.:. ..|.. : .:
. ::
. |...:...: .
. |: ::. ::.
. |...: .:
. |: ::. ::
.:. ..|...:...: .
.:. .| .:|.:|: ::. ::.
. . .:
.:. .:..::
.:. ::. :: .
.:. .|..:| .
.:. ::. ::
.:. .:..::
.:. ::. ::.. .: .
.:. .:|: :. ::.
. | .:
. |: ::. ::
. |...:...: .
. |: ::. ::.
. |...: .:
.:. .:|: ::. ::
.:. :: ::.. .: .
. .
.:. ::. ::
.:. .:..::
.:. ::. :: .
.:. .|..:| .
.:. .. .:
. |: :. ::
. ..: .
. |: ::. ::.
. |...: .:
. |: ::. ::
.:. ..|...:...: .
.:. .| .:|.:|: ::. ::.
============================================ ===================================================================================================================================
| |
| |
| |
| |
| |
| |
| "I'm glad you're ___" |
| |
| |
| |
| 46.3% |
| |
| |
| |
| //===\. |
| | xx -------------------------------------------------- 37.0% |
| ||| |
| #|| |
| ||| |
| | x/:: |
| \\-=-x: ---=---=- 6.5% |
| |
| /x=--\\ |
| | xx\| |
| ||: |
| ||| #=## 3.0% |
| |:| |
| | ///| |
| \\=--// |
| x/---\\ |
| | x\\. === 2.0% |
| ||: |
| ||: |
| ||: |
| | #/:: |
| \\-=-:: -- 1.5% |
| //==-\\ |
| : x\\. |
| ||| |
| ||| |
| ||: -= 1.2% |
| | #xx: |
| /x=-=\: \\==-:: |
| | xx\: |
| || |
| #|| # 1.0% |
| ||: |
| | x//| |
| \x==-/: |
| |
| |
| |
| |
================================================================================================================================================================================
|
|
%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@
#=#=#=#=#=#=#=#=#=#=#=#=#=#=#=#
## @#
%| %#
## @#
%| %#
## @#
%| %#
## @#
%| %#
## @#
%| %#
## @#
%| : %#
## xxxx=x @# interested
%| #|%%%%x# %#
## |#@###|# @#
%| |##%%#xx %#
## \%x%=# @#
%| %# okay
## : @#
%| interested %#
## #x%#%x @#
%| #|###=#\ %#
## #%#%@%%| @# following
%| |#%#%#%/ %#
## ####=# @#
%| %#
## : @#
%| #xxx#\ okay %# excited
## |%%%@=## @#
%| |%%#|#|# %#
## #|#%%%xx @#
%| \x##=x %#
## @# listening
%| : %#
## following x##x=x @#
%| |##x%##x %#
## #%#%@%%| @#
%| ##%%##xx %# here
## #x#x=x @#
%| %#
## xxx### : @#
%| ||%#%=xx excited %#
## |#@###|# @# safe
%| ####%%xx %#
## x%x#=# @#
%| %#
## : @#
%#%%%#%%here%#%%%#%%%#%%%#%%%#%
###############################
+ +
/==-
x###=xx\
|%####xx||
####@#%||| concerned @@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@%@@@
|#%%%%#x||
##x%-#//
\-#=
concerned
+ +
/ \:
| \.
|
|
|
| /.
x //
· 9 min
Hyperparameters in the prompting process
Temperature and Top P are usually explained as “randomness”. What they actually do to the words a model chooses between, and why temperature 0 doesn’t always give the best answers.
9.[Rolle]:9b 9.[Rolle]:9c 3.[Begrep]:3c 3.[Rolle]:3a 5.[Rolle]:5a 9.[Mønster]:9c 1.[Eksempel]:1c 8.[Vekt]:8c 3.[Format]:3c 9.[Kontekst]:9a 1.[Kontekst]:1a 1.[Beg
8.[Token]:8c 9.[Begrep]:9a 5.[Eksempe ]:5b 2.[Kontekst]:2b 5.[Eksempel]:5b 2.[Tone]:2c 6.[Struktur]:6a 5.[Tone]:5a 2.[Mønster]:2b 5.[Mønster]:5b 1 [Mønster]:1c 4
8.[Kontekst]:8b 7.[Format]:7b 4.[Mønster]:4c 6.[Struktur]:6a 6.[Struktur]:6a 2.[Vekt]:2a 2.[Begrep]:2a 8.[Kontekst :8b 9.[Relasjon]:9a 9.[ olle]:9a 3.[Format]:3b
7.[Eks ]:7a 5.[Kontekst]:5c 1.[Strukt [Eksempel]:2a 4.[Relasjon]:4b 5b 2.[Mønster]:2a 4.[Begrep]
6.[Fagfelt :6b 8.[Fagfelt]:8c n]:7a .[Ek empe ] p]:4c 3.[Begrep] sem el]: a 7 [Eks Fagfelt]:2b 2.[Konteks
9.[Konteks ]:9b 7.[ egre kt]:8 one]:3 3. Møns er]: a 9 [Rol e]:9 lt]:2 eg ep]: c 2 [Kon ekst :2b 8.[V kt]: tekst 6.[Str
3.[Fagfelt :3a 3.[ ken] 3b . 1 [Rel sjon :1a 6.[R lle] 6c .[St uktu ]:1a ]:2 5. Møns er]: b 2 [Rel sjon : a 9.[R lle] mat]:1c 3.
2.[Token]: c 2 [Rol e]:2 8. kst teks ]:7b 1.[ agfe t]:1 2. Møns er] .[E 6.[V kt]: b 8 [For at]: b 9 [Rol e] .[T 2.[Str
4.[Konteks ]:4c 8.[M c 8.[Eksempel] ks ]:6b 4.[ elas on]: elasjon]:6c 8.[ [To en]: c 4 [Str ktu :4a 4.[V ]:4c
4.[Tok n]:8b 2.[Kontekst]:2b 2.[R [ ekt]:7c 8.[Tone]:8c 8.[Tok felt] 1c 5.[Format]:5c 5.[Tok
1.[Fagfelt]:1c 7.[Relasjon]:7b 1.[Begrep]:1b 3.[Begrep .[Eksempel]:2b 9.[Mønster]:9c 6.[Eksempel]:6a 4.[E l]:4a 5.[Rolle]:5a 1.[Format]:1b 1.[Tone]:1a
5.[Begrep]:5b 9.[Tone]:9c 3.[Fagfelt]:3b 4.[Fo mat]:4a 7.[Vekt]:7a 8.[Rolle]:8a 1.[Eksempel]:1b 9.[Tone]:9b 6.[Rolle]:6c 2.[Begrep]:2c 2.[Eksempel]:2a 7.[Begrep]
9.[Struktur]:9b 1.[Struktur]:1b 2.[Tone]:2a 6.[Rolle]:6c 5.[Vekt]:5b 4.[Kontekst]:4a 1.[Vekt]:1a 6.[Tone]:6c 8.[Relasjon]:8a 9.[Tone]:9c 8.[Format]:8b 2.[Fagfelt
: -
.. :- :.: .::. : . : :-
.. :. . :. :
. :. . :. .-
\ . : . . .
: :: : . . :
. /. \ :: .
. : . : /
. .. : :. : .
:: .. /:
: :
: : /
. . :
.: : /.
:. :
/. : : . .: \
.
\ . : .
. :. /
::. : .:. : /:
-.. . : . : -.
-. . -.: : -:.
:. .: ... .
--:: . : .- -:.:: : .... -..-:.
================================================================================================================================================================================
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| -==# = -- =-==-= # -- :-=#=- = -- |
| =- ---=-==- :-=====- |
| ===#=- :=--=-----=--#= -:-==---===--=# |
| ----===- /=---=-----=-==-=== :=-=---------==--#= |
| ----:::=--- :/:--=---:-: .-:.--::==- /:x:=---=--: --.:--.-=-- |
| --# //x/x--:- -== /////=--- -:- |
| | | .--- .::::::- | | :--- :://://: | | ---- |
| | | -\\ /x/x// | | | -\\ ///:// | | | -xx / |
| | | | :\# ///::. | | | | | \\x .:::// | | | | \\\ //: |
| | | | :x ::::/ | | | | | \\ /x//: | | | | | \x ::// |
| | | | | xx ://// | | | | | :# ::::: | | | | | | \x ////. |
| | | | | \x :////: | | | | | | | :x .:://: | | | | | | :\ ::::// |
| | | | | \x //:: | | | | | | | \\ //// | | | | | | | :# :/// |
| | | | | \x | | | | | | | \\ | | | | | | | | :x |
| | | | | \\ | | | | | | | | :# | | | | | | | | \\ |
| | | | | \x | | | | | | | | :x | | | | | | | | \\ |
| | | | | ::/\# | | | | | | | /::\x | | | | | | | | //::x |
| | | | | .//// \x | | | | | | | //:// :\ | | | | | | | .//:/ :\ |
| | | | | ///:. \x | | | | | | | /::// :\ | | | | | | | ://:: \\ |
| | | | | ///:: :\ | | | | | | //:// \\ | | | | | ////: \x |
| | | | //:::/ \x | | | | | ://:/: \x | | | | | /////: :\ |
| | | | /:/:// xx = | | | | ////:/ \x =| | | | ////:/ x\ |
| | | ::/:///: :\x - | | | .////::/ \\\ = | | | :/x://: :x\ |
| | | ::--x//// \x - | | | :---///:/ \\ - | | --=-//x/: \ |
| : :---===-//. .\xx =:- --------//: :\\x =-: :::---:-//. |
| ---==#==----:: :xx ==#-==-=-----/ \\\ =#---======-:: |
| =#=---:---- :\-- =-#--=----: \\-# -===--:-::: |
| #=-----===- :--=-== #=-==-=---: ::--=== =-=#-----=- |
| -----=--:: -:----=--=--=:-: .:-===--=-:---:: |
| - :. :- |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
================================================================================================================================================================================
==-= = #-== - =#=- #
=-=#%==#%==#= #=-###=#%#=## =#=-##==%#=-=
-=#==%#=#%\x -=#===%#=##\ -=#==##%##x#
=##-#%##@### -=##=#%#x@x# ==#===##%##\
#==##x@##% -##==@x%%# ###==x##@%
==##x%%#%@ -=%x#%#@#% =##%%%###@
x#x#%##@# ##x%#%#%x \xx#%@#%#
#x#%##@% #x#%%%%# %##x@%#@
x#%%x%@# #x%#%%#% x%##@#@#
xx%%#%@@# #x@#%%#@@ xx%#%%@#%
x#@##%@% \%x%%#%@ x#%#%%#%
xx@%##@@% \%x%@%#@% #x@#%@##%
\#%%##%@ #x#%%### \%x%@%##
\x%@%##%# #x#%@%%## \%##@@%%#
xx%%@##x xx#%@@%% #x##@@@%
xx#%@@@@% ##x#%@%%# ###x####%
xxx##%%% ###x##%% #%%%%@%#
#%#####%# #%@@@@@## x##%##%@x
#%@%@@@#x \#x#x%@#x xx#%%#@
xx#x###%@# x#%@%##@x# ###x@#
x##%%%#### xx##@#%##\ x%##
#x#x%@%x% #%x%%#%%x x%
#%#xx@### ##x%##%=# struktur
\x#%x#%=%=# relasjon \#x%x#@#===
kontekst \##=###=#== x##=#%%#%##
##= x#####=%=##-=# ##===#====##==
-=== ==#==#==#=== ===#=##=-=#=
# -#=- -==#
mønster begrep
+ +
|
| |
| |
| |
| |
| |
| |
| | |
@ @ @ @ @
| | | | |
| | | |
| |
| |
| |
| |
| |
| |
|
+ +
*
.
* +
*
* *
+
+ .
*
.
+ +
+
.
* +
+ +
+
.
+ +
* .
*
+ *
+
.
· 8 min
Prompt engineering: It’s not about instructions, it’s about context
Why language models don’t follow instructions the way a computer follows code, and why good prompting is about building the right context.