I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
,
p
p
.
953
~
961
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
1
.
p
p
9
5
3
-
9
6
1
953
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Sca
la
ble resume s
creening
using
lar
g
e lang
ua
g
e mo
del
M
eta
AI
v
ersio
n 3
Asm
it
a
Deshm
uk
h
1,
2
,
Anj
a
li
Ra
ut
1
,
Veda
nt
Deshm
uk
h
3
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
H
V
P
M
C
o
l
l
e
g
e
o
f
En
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
,
A
mr
a
v
a
t
i
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
En
g
i
n
e
e
r
i
n
g
,
K
.
J.
S
o
m
a
i
y
a
I
n
st
i
t
u
t
e
o
f
T
e
c
h
n
o
l
o
g
y
,
M
u
mb
a
i
,
I
n
d
i
a
3
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
En
g
i
n
e
e
r
i
n
g
,
S
a
r
d
a
r
P
a
t
e
l
I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
M
u
m
b
a
i
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 5
,
2
0
2
5
R
ev
is
ed
Dec
3
0
,
2
0
2
5
Acc
ep
ted
J
an
2
2
,
2
0
2
6
Th
is
re
se
a
rc
h
p
a
p
e
r
e
x
p
l
o
re
s
th
e
u
se
o
f
larg
e
lan
g
u
a
g
e
m
o
d
e
l
M
e
ta
AI
3
(
LL
a
M
A
3
)
f
o
r
a
u
t
o
m
a
ti
n
g
th
e
re
su
m
e
sc
re
e
n
in
g
p
r
o
c
e
ss
.
Trad
it
io
n
a
l
re
su
m
e
sc
re
e
n
in
g
m
e
th
o
d
s
th
a
t
re
l
y
o
n
k
e
y
wo
rd
se
a
rc
h
i
n
g
a
n
d
h
u
m
a
n
re
v
i
e
w
c
a
n
b
e
in
e
fficie
n
t,
b
ias
e
d
,
a
n
d
fa
il
to
id
e
n
ti
f
y
q
u
a
li
fied
c
a
n
d
i
d
a
tes
.
L
LaMA
3
,
train
e
d
o
n
larg
e
-
sc
a
le
tex
t
d
a
tas
e
ts,
h
a
s
th
e
p
o
ten
ti
a
l
to
a
c
c
u
ra
tely
a
n
a
ly
z
e
re
su
m
e
s
b
y
u
n
d
e
rsta
n
d
in
g
c
o
n
t
e
x
t
a
n
d
se
m
a
n
ti
c
d
e
tails
b
e
y
o
n
d
sim
p
le
k
e
y
wo
r
d
m
a
tch
in
g
.
T
h
e
stu
d
y
p
re
se
n
ts
a
sy
ste
m
th
a
t
c
o
n
v
e
rts
re
su
m
e
P
DFs
to
tex
t
,
in
p
u
ts
th
e
tex
t
a
lo
n
g
wi
th
th
e
j
o
b
d
e
sc
rip
ti
o
n
in
t
o
th
e
LL
a
M
A
3
m
o
d
e
l,
a
n
d
g
e
n
e
ra
tes
a
ra
n
k
e
d
li
s
t
o
f
c
a
n
d
i
d
a
tes
with
re
a
so
n
in
g
fo
r
th
e
ir
jo
b
fit
.
T
h
is
d
isc
u
ss
e
s
t
h
e
d
a
ta
p
r
e
p
a
ra
ti
o
n
,
m
o
d
e
l
se
tu
p
,
a
n
d
p
e
rfo
rm
a
n
c
e
e
v
a
lu
a
ti
o
n
o
f
th
is
sy
ste
m
.
Re
su
lt
s
sh
o
w
LL
a
M
A
3
c
a
n
ra
p
id
l
y
p
ro
c
e
ss
b
a
tch
e
s
o
f
re
su
m
e
s
wh
il
e
re
d
u
c
in
g
h
u
m
a
n
b
ias
in
th
e
sc
re
e
n
in
g
p
r
o
c
e
ss
.
Th
e
sy
ste
m
a
ims
to
stre
a
m
li
n
e
h
ir
in
g
b
y
a
u
to
m
a
ti
n
g
th
e
in
it
ial
re
su
m
e
sc
re
e
n
in
g
sta
g
e
to
su
rfa
c
e
to
p
c
a
n
d
id
a
tes
fo
r
fu
rt
h
e
r
in
-
d
e
p
th
e
v
a
lu
a
ti
o
n
.
K
e
y
b
e
n
e
fit
s
in
c
lu
d
e
imp
ro
v
e
d
a
c
c
u
ra
c
y
in
id
e
n
ti
f
y
in
g
re
lev
a
n
t
sk
il
ls,
re
d
u
c
e
d
b
ias
c
o
m
p
a
r
e
d
to
h
u
m
a
n
sc
re
e
n
e
rs,
a
n
d
sig
n
ifi
c
a
n
t
t
ime
sa
v
in
g
s
fo
r
r
e
c
ru
it
e
rs.
Th
e
p
a
p
e
r
a
lso
e
x
a
m
in
e
s
e
th
ica
l
c
o
n
sid
e
ra
ti
o
n
s
a
ro
u
n
d
u
sin
g
AI
fo
r
h
iri
n
g
d
e
c
isio
n
s.
Ov
e
ra
ll
,
th
is
wo
rk
d
e
m
o
n
stra
tes
t
h
e
p
r
o
m
isin
g
a
p
p
li
c
a
ti
o
n
o
f
larg
e
lan
g
u
a
g
e
m
o
d
e
ls
(
LLM
s
)
li
k
e
L
LaMA
3
to
tran
sfo
rm
a
n
d
e
n
h
a
n
c
e
re
su
m
e
sc
re
e
n
in
g
p
ra
c
ti
c
e
s.
K
ey
w
o
r
d
s
:
Au
to
m
ated
r
ec
r
u
itm
en
t
L
ar
g
e
lan
g
u
ag
e
m
o
d
els
L
L
aM
A
3
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
R
esu
m
e
s
cr
ee
n
in
g
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Asmita
Desh
m
u
k
h
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
an
d
E
n
g
in
ee
r
in
g
,
HVPM
C
o
lleg
e
o
f
E
n
g
i
n
ee
r
in
g
an
d
T
ec
h
n
o
lo
g
y
Am
r
av
ati,
Ma
h
ar
ash
tr
a,
I
n
d
ia
E
m
ail: a
s
m
itad
esh
m
u
k
h
7
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
Au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
h
as
em
er
g
ed
as
a
cr
itical
ar
ea
o
f
in
ter
est
in
r
ec
r
u
itm
en
t,
aim
in
g
to
ad
d
r
ess
th
e
in
ef
f
icien
cies
an
d
b
iases
in
h
er
en
t
in
tr
ad
itio
n
al
m
an
u
al
r
ev
iew
m
eth
o
d
s
.
A
p
i
o
n
ee
r
in
g
s
y
s
tem
f
o
r
ex
tr
ac
tin
g
in
f
o
r
m
ati
o
n
f
r
o
m
u
n
s
tr
u
ctu
r
ed
r
esu
m
es
u
s
in
g
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
tech
n
iq
u
es
m
ar
k
ed
a
n
in
itial
s
tep
to
war
d
s
tr
ea
m
lin
in
g
th
e
s
cr
ee
n
in
g
p
r
o
ce
s
s
[
1
]
.
T
h
e
ef
f
ec
tiv
e
u
s
e
o
f
r
u
le
-
b
ased
NL
P
to
ex
tr
ac
t
h
o
u
s
in
g
s
tatu
s
f
r
o
m
clin
ical
n
ar
r
ativ
es
h
ig
h
lig
h
te
d
th
e
p
o
ten
tial
o
f
u
tili
zin
g
co
n
tex
tu
al
c
u
es
in
u
n
s
tr
u
ctu
r
ed
tex
t
f
o
r
id
en
tif
y
i
n
g
s
o
cial
d
eter
m
in
an
ts
o
f
h
ea
l
th
[
2
]
.
T
h
e
u
s
e
o
f
NL
P
f
o
r
p
r
e
d
ictiv
e
m
o
d
elin
g
o
f
HI
V
d
iag
n
o
s
es
h
ig
h
lig
h
ts
th
e
ca
p
ab
ilit
y
o
f
th
ese
tech
n
iq
u
es
to
ex
tr
ac
t
m
ea
n
in
g
f
u
l
in
f
o
r
m
a
tio
n
f
r
o
m
co
m
p
le
x
tex
tu
al
d
ata
[
3
]
.
T
h
e
v
e
r
s
atility
o
f
NL
P
ac
r
o
s
s
d
o
m
ain
s
s
h
o
ws
its
u
s
e
f
o
r
ac
cu
r
ately
ca
lcu
latin
g
ad
en
o
m
a
d
etec
tio
n
r
ates
in
s
cr
ee
n
in
g
co
lo
n
o
s
co
p
ies
an
d
f
o
r
lar
g
e
-
s
ca
le
lab
elin
g
o
f
clin
ical
r
ec
o
r
d
s
[
4
]
,
[
5
]
.
T
h
ese
ap
p
r
o
ac
h
es
r
elian
ce
o
n
d
o
m
a
in
-
s
p
ec
if
ic,
r
u
le
-
b
ased
,
o
r
tr
a
d
itio
n
al
tech
n
iq
u
es
lim
its
th
e
ab
ilit
y
to
ca
p
tu
r
e
n
u
an
ce
d
c
o
n
tex
t
an
d
g
en
e
r
alize
to
d
iv
er
s
e,
u
n
s
tr
u
ctu
r
ed
d
ata
lik
e
r
esu
m
es.
T
h
e
in
teg
r
atio
n
o
f
m
ac
h
in
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
953
-
9
6
1
954
lear
n
in
g
alg
o
r
ith
m
s
s
u
ch
as
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM)
an
d
r
an
d
o
m
f
o
r
est
(
R
F)
h
as
en
h
an
ce
d
r
esu
m
e
s
cr
ee
n
in
g
b
y
im
p
r
o
v
in
g
th
e
a
cc
u
r
ac
y
an
d
s
ca
lab
ilit
y
o
f
tex
t
class
if
icatio
n
.
A
two
-
lev
el
class
if
icatio
n
s
y
s
te
m
was
ap
p
lied
f
o
r
ex
tr
ac
tin
g
f
e
atu
r
es
f
r
o
m
r
esu
m
es,
ac
h
ie
v
in
g
8
5
%
ac
cu
r
ac
y
in
h
ea
d
in
g
p
r
ed
ictio
n
u
s
in
g
th
e
ex
tr
em
e
g
r
a
d
ien
t
b
o
o
s
tin
g
(
X
GB
o
o
s
t
)
alg
o
r
ith
m
[
6
]
.
A
c
o
s
in
e
s
im
ilar
ity
-
b
ased
s
y
s
tem
,
a
s
p
r
esen
ted
in
[
7
]
,
d
em
o
n
s
tr
ated
h
ig
h
ac
cu
r
ac
y
an
d
ef
f
icien
cy
in
e
x
tr
ac
tin
g
r
elev
an
t
s
k
ills
an
d
q
u
alif
icatio
n
s
.
An
SVM
-
b
ased
m
o
d
el
im
p
r
o
v
ed
h
ir
in
g
ef
f
icien
cy
b
y
m
atch
in
g
r
esu
m
es
wit
h
jo
b
d
escr
ip
tio
n
s
[
8
]
,
wh
ile
a
th
r
ee
-
p
h
ase
NL
P
-
d
r
iv
en
a
p
p
r
o
ac
h
en
h
an
ce
d
s
em
an
tic
r
esu
m
e
an
aly
s
is
an
d
ca
n
d
id
ate
r
an
k
in
g
[
9
]
.
T
h
e
u
s
e
o
f
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
an
d
SVM
was
ex
p
lo
r
ed
f
o
r
class
if
y
in
g
ca
n
d
id
ates
an
d
r
esu
m
es
[
1
0
]
.
KNN
class
if
ied
ca
n
d
id
ates
in
to
d
if
f
e
r
en
t
p
ac
k
ag
es
b
ased
o
n
th
eir
p
r
o
f
iles
,
wh
ile
SVM
ca
teg
o
r
ized
r
esu
m
es
in
to
test
in
g
an
d
d
ev
elo
p
m
e
n
t
p
r
o
f
iles
.
Ma
c
h
in
e
lear
n
in
g
tech
n
iq
u
es
f
o
r
r
esu
m
e
s
cr
ee
n
in
g
ar
e
lim
ited
b
y
t
h
e
n
ee
d
f
o
r
m
an
u
al
f
ea
tu
r
e
ex
tr
ac
tio
n
,
p
o
o
r
h
an
d
lin
g
o
f
u
n
s
tr
u
ctu
r
ed
r
esu
m
es,
an
d
d
if
f
ic
u
lty
in
p
r
o
ce
s
s
in
g
lo
n
g
-
f
o
r
m
tex
t
an
d
co
n
tex
t.
A
d
ee
p
r
esid
u
al
co
n
v
o
lu
ti
o
n
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
DR
-
C
L
STM
)
n
etwo
r
k
co
m
b
in
es
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN
)
an
d
lo
n
g
s
h
o
r
t
-
te
r
m
m
em
o
r
y
(
L
STM
)
to
p
r
e
d
ict
o
p
tio
n
p
r
ices
m
o
r
e
ac
c
u
r
ately
an
d
ef
f
icien
tly
th
an
tr
ad
itio
n
al
m
o
d
els
li
k
e
B
lack
-
Sch
o
les
b
y
ca
p
tu
r
i
n
g
c
o
m
p
lex
m
ar
k
et
d
y
n
am
ics
[
1
1
]
.
T
h
e
s
tu
d
y
[
1
2
]
ex
p
lo
r
es
NL
P
an
d
L
STM
-
b
ased
m
o
d
els,
s
h
o
wca
s
in
g
im
p
r
o
v
em
en
ts
in
r
esu
m
e
class
if
icatio
n
ac
cu
r
ac
y
ac
r
o
s
s
v
ar
io
u
s
ap
p
licatio
n
s
.
C
NN
an
d
L
STM
wer
e
u
s
ed
to
d
iag
n
o
s
e
ar
r
h
y
th
m
ias
f
r
o
m
E
lectr
o
ca
r
d
io
g
r
am
(
E
C
G
)
s
ig
n
als
[
1
3
]
,
wh
ile
co
m
b
in
in
g
Fo
u
r
ier
-
B
ess
el
ex
p
an
s
io
n
with
L
STM
th
at
ac
h
iev
ed
9
0
.
0
7
%
class
if
icatio
n
ac
cu
r
a
cy
[
1
4
]
.
A
m
u
lti
-
lay
er
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
(
R
NN)
im
p
r
o
v
es
s
k
ill
an
d
ex
p
er
ien
ce
id
e
n
tific
atio
n
wh
ile
in
teg
r
atin
g
en
g
a
g
em
en
t
s
u
r
v
ey
s
an
d
b
u
s
in
ess
in
tellig
en
ce
to
s
u
p
p
o
r
t
h
u
m
a
n
r
eso
u
r
ce
s
(
HR
)
d
ec
is
io
n
s
an
d
t
ea
m
allo
ca
tio
n
[
1
5
]
.
Dee
p
lear
n
in
g
m
o
d
els
f
o
r
r
es
u
m
e
s
cr
ee
n
in
g
h
as
lim
itatio
n
s
s
u
ch
as
th
e
n
ee
d
f
o
r
p
r
e
-
p
r
o
c
ess
in
g
an
d
em
b
ed
d
in
g
s
,
f
ix
e
d
c
o
n
tex
t
win
d
o
ws
th
at
h
in
d
er
lo
n
g
r
esu
m
e
h
an
d
lin
g
,
an
d
d
if
f
icu
lty
ca
p
tu
r
in
g
co
m
p
le
x
cr
o
s
s
-
p
ar
ag
r
ap
h
r
elatio
n
s
h
ip
s
.
T
r
an
s
f
o
r
m
e
r
s
,
p
ar
tic
u
lar
ly
b
id
ir
ec
tio
n
al
en
c
o
d
er
r
e
p
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
,
h
av
e
r
ev
o
lu
tio
n
ize
d
r
esu
m
e
s
cr
ee
n
in
g
with
th
eir
ab
ilit
y
to
ca
p
tu
r
e
b
id
ir
ec
tio
n
al
co
n
tex
t.
T
h
e
s
tu
d
y
[
1
6
]
u
s
es
B
E
R
T
an
d
n
am
ed
en
tity
r
e
co
g
n
itio
n
(
NE
R
)
to
au
to
m
ate
r
esu
m
e
s
cr
ee
n
in
g
,
im
p
r
o
v
in
g
ac
cu
r
ac
y
a
n
d
r
ed
u
cin
g
m
an
u
al
ef
f
o
r
t.
A
B
E
R
T
-
b
ased
f
r
am
ewo
r
k
f
in
e
tu
n
e
d
o
n
h
is
to
r
ical
jo
b
ap
p
licatio
n
d
ata
is
u
s
ed
to
o
b
jectiv
ely
p
r
e
d
ict
p
e
r
s
o
n
-
jo
b
f
i
t
an
d
im
p
r
o
v
e
r
esu
m
e
s
cr
ee
n
i
n
g
e
f
f
icien
cy
[
1
7
]
.
Sen
ten
ce
-
B
E
R
T
(
S
-
B
E
R
T
)
is
wid
ely
u
s
ed
in
au
to
m
ated
r
e
s
u
m
e
s
cr
ee
n
in
g
f
o
r
g
e
n
er
atin
g
em
b
ed
d
in
g
s
an
d
im
p
r
o
v
in
g
r
elev
a
n
ce
r
an
k
i
n
g
v
ia
co
s
in
e
s
im
ilar
ity
,
o
f
ten
o
u
tp
er
f
o
r
m
i
n
g
tr
ad
itio
n
al
B
E
R
T
m
o
d
els
in
ac
cu
r
ac
y
an
d
ef
f
icien
c
y
th
r
o
u
g
h
h
y
b
r
i
d
NL
P
ap
p
r
o
ac
h
es
[
1
8
]
–
[
2
0
]
.
Sen
ten
ce
-
p
air
B
E
R
T
(
SP
B
E
R
T
)
,
a
f
in
e
-
tu
n
ed
B
E
R
T
v
ar
ian
t,
en
c
o
d
es
r
esu
m
es
an
d
jo
b
p
o
s
tin
g
s
i
n
to
u
n
i
f
ied
em
b
ed
d
in
g
s
to
p
r
e
d
ict
c
o
m
p
atib
ilit
y
th
r
o
u
g
h
ad
ap
ted
n
ex
t
-
s
en
ten
ce
p
r
ed
ict
io
n
[
2
1
]
.
T
r
an
s
f
o
r
m
er
m
o
d
el
is
a
p
o
wer
f
u
l
tech
n
iq
u
e
f
o
r
r
esu
m
e
s
cr
ee
n
i
n
g
,
b
u
t
ar
e
lim
ited
b
y
ch
allen
g
es
in
ze
r
o
o
r
f
ew
s
h
o
t
class
if
icatio
n
,
h
a
n
d
lin
g
lo
n
g
tex
ts
,
an
d
r
eq
u
ir
in
g
co
s
tly
f
in
e
-
tu
n
in
g
.
L
ar
g
e
lan
g
u
a
g
e
m
o
d
els
(
L
L
Ms)
lik
e
g
en
er
ativ
e
p
r
e
-
tr
ain
ed
tr
an
s
f
o
r
m
er
s
(
GPT
)
,
clau
d
e
,
an
d
lar
g
e
lan
g
u
ag
e
m
o
d
el
M
eta
AI
2
(
L
L
aM
A
2
)
a
r
e
in
cr
ea
s
in
g
l
y
u
s
ed
in
ab
s
tr
ac
t
s
cr
ee
n
in
g
to
a
u
to
m
ate
ev
alu
atio
n
th
r
o
u
g
h
ad
v
an
ce
d
lan
g
u
a
g
e
u
n
d
er
s
tan
d
i
n
g
an
d
co
n
tex
tu
a
l
an
aly
s
is
[
2
2
]
.
T
h
e
p
ap
er
[
2
3
]
p
r
o
p
o
s
es
an
LLM
-
b
ased
r
esu
m
e
s
cr
ee
n
in
ag
en
t
f
r
a
m
ewo
r
k
f
o
r
r
esu
m
e
s
cr
ee
n
in
g
t
h
at
is
1
1
tim
es
f
aster
th
an
m
a
n
u
al
m
eth
o
d
s
a
n
d
ac
h
iev
es
an
8
7
.
7
3
%
F1
-
s
co
r
e,
s
u
r
p
ass
in
g
G
PT
-
3
.
5
i
n
s
u
m
m
a
r
izatio
n
,
g
r
a
d
in
g
,
an
d
d
ec
is
io
n
-
m
ak
in
g
.
L
L
M
f
r
am
ewo
r
k
with
r
etr
iev
al
-
au
g
m
en
ted
g
en
er
at
io
n
(
R
AG)
au
to
m
ates
r
esu
m
e
s
cr
ee
n
in
g
th
r
o
u
g
h
co
n
tex
t
-
awa
r
e
ex
tr
ac
tio
n
,
ev
al
u
atio
n
,
s
u
m
m
ar
izatio
n
,
an
d
s
c
o
r
in
g
,
clo
s
ely
alig
n
in
g
with
H
R
ass
es
s
m
en
ts
an
d
en
h
an
cin
g
r
ec
r
u
itm
en
t scala
b
ilit
y
[
2
4
]
,
[
2
5
]
.
C
u
r
r
en
t
r
esu
m
e
s
cr
ee
n
in
g
m
et
h
o
d
in
clu
d
in
g
r
u
le
-
b
ased
s
y
s
tem
s
,
class
ica
l
m
ac
h
in
e
lear
n
in
g
m
o
d
els,
an
d
ea
r
ly
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
es a
r
e
lim
ited
in
h
an
d
lin
g
u
n
s
tr
u
ctu
r
ed
a
n
d
len
g
th
y
r
esu
m
es d
u
e
to
th
eir
lack
o
f
co
n
tex
t
u
al
u
n
d
er
s
tan
d
in
g
,
d
ep
en
d
e
n
ce
o
n
m
an
u
al
f
ea
t
u
r
e
ex
tr
ac
tio
n
,
r
estricte
d
co
n
tex
t
win
d
o
ws,
an
d
h
ig
h
f
in
e
-
tu
n
in
g
co
s
ts
.
T
o
o
v
er
co
m
e
th
ese
co
n
s
tr
ain
ts
,
th
er
e
i
s
a
g
r
o
win
g
n
ee
d
f
o
r
a
s
ca
lab
le,
ad
ap
tab
le,
an
d
co
n
tex
t
-
awa
r
e
s
o
lu
tio
n
ca
p
ab
le
o
f
ac
cu
r
ately
e
v
alu
atin
g
r
esu
m
es
with
m
in
im
al
m
an
u
a
l
in
p
u
t.
T
h
is
s
tu
d
y
s
ee
k
s
to
en
h
an
ce
r
esu
m
e
s
cr
ee
n
in
g
b
y
u
tili
zin
g
L
L
aM
A
3
m
o
d
el
’
s
ca
p
ab
ilit
ies
to
r
ep
lace
b
iased
an
d
in
ef
f
icien
t
k
ey
wo
r
d
-
b
ased
m
eth
o
d
s
,
en
ab
lin
g
ac
cu
r
ate
ca
n
d
id
ate
r
an
k
in
g
,
r
ed
u
ce
d
h
u
m
an
b
ias,
an
d
au
to
m
atio
n
o
f
th
e
in
itial st
ag
e
s
o
f
h
ir
in
g
.
T
h
e
r
em
a
i
n
d
er
o
f
th
i
s
p
ap
e
r
is
o
r
g
an
i
z
ed
a
s
f
o
l
lo
w
s
:
se
c
t
i
o
n
2
d
e
t
a
i
l
s
t
h
e
m
e
th
o
d
o
lo
g
y
,
i
n
c
l
u
d
i
n
g
t
h
e
d
a
ta
p
r
e
p
ar
a
t
io
n
p
r
o
c
e
s
s
,
P
D
F
-
to
-
te
x
t
co
n
v
e
r
s
io
n
,
s
y
s
t
e
m
c
o
n
f
i
g
u
r
a
t
i
o
n
u
s
i
n
g
L
M
S
t
u
d
i
o
,
an
d
th
e
d
e
s
i
g
n
o
f
t
h
e
L
3
-
b
a
s
ed
r
e
s
u
m
e
s
c
r
e
en
i
n
g
f
r
a
m
e
w
o
r
k
.
S
ec
t
i
o
n
3
p
r
e
s
en
t
s
t
h
e
r
e
s
u
l
t
s
f
o
l
l
o
w
ed
b
y
s
e
c
t
io
n
4
d
i
s
c
u
s
s
i
o
n
,
c
o
v
e
r
in
g
m
o
d
e
l
c
o
n
f
ig
u
r
a
t
i
o
n
,
p
r
o
c
e
s
s
i
n
g
e
f
f
i
c
i
en
cy
,
c
o
m
p
ar
a
t
i
v
e
an
al
y
s
i
s
o
f
c
a
n
d
i
d
a
te
e
v
a
l
u
a
t
i
o
n
s
,
an
d
p
er
f
o
r
m
a
n
c
e
b
en
ch
m
ar
k
in
g
a
g
a
in
s
t
o
t
h
e
r
L
L
M
s
.
S
e
c
t
i
o
n
5
co
n
c
l
u
d
e
s
t
h
e
s
t
u
d
y
,
s
u
m
m
ar
i
z
i
n
g
k
e
y
f
i
n
d
i
n
g
s
,
s
o
c
i
e
t
a
l
im
p
l
i
c
a
t
io
n
s
,
a
n
d
f
u
tu
r
e
s
co
p
e
w
i
t
h
r
e
c
o
m
m
en
d
a
ti
o
n
s
f
o
r
e
t
h
i
c
a
l
A
I
d
e
p
lo
y
m
e
n
t
i
n
r
e
cr
u
i
t
m
e
n
t
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ca
la
b
le
r
esu
me
s
creen
in
g
u
s
i
n
g
la
r
g
e
la
n
g
u
a
g
e
mo
d
el
Meta
A
I
ve
r
s
io
n
3
(
A
s
mita
Desh
mu
kh
)
955
2.
M
E
T
H
O
D
T
h
is
s
ec
tio
n
o
u
tlin
es
th
e
m
e
th
o
d
o
lo
g
ical
f
r
am
ew
o
r
k
ad
o
p
ted
to
d
ev
elo
p
an
au
to
m
at
ed
r
esu
m
e
s
cr
ee
n
in
g
s
y
s
tem
u
s
in
g
th
e
L
L
aM
A
3
lan
g
u
ag
e
m
o
d
el.
I
t
in
clu
d
es
th
e
f
o
r
m
al
p
r
o
b
le
m
f
o
r
m
u
latio
n
,
d
ata
p
r
ep
ar
atio
n
,
s
y
s
tem
ar
ch
itectu
r
e,
m
o
d
el
in
teg
r
atio
n
,
a
n
d
c
o
n
f
ig
u
r
atio
n
s
ettin
g
s
.
T
h
e
a
p
p
r
o
ac
h
em
p
h
asizes
th
e
u
s
e
o
f
co
n
tex
tu
al
em
b
ed
d
i
n
g
s
an
d
p
r
o
m
p
t
-
b
ased
r
ea
s
o
n
in
g
f
o
r
ac
c
u
r
ate
ca
n
d
id
ate
-
jo
b
m
atch
in
g
,
wh
ile
en
s
u
r
in
g
f
ai
r
n
ess
,
s
ca
lab
ilit
y
,
an
d
ex
p
lain
ab
ilit
y
.
2
.
1
.
P
r
o
blem
def
ini
t
io
n
Au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
ca
n
b
e
f
o
r
m
ally
m
o
d
eled
as
a
m
u
lti
-
cr
iter
io
n
r
an
k
in
g
p
r
o
b
l
em
,
wh
er
e
th
e
o
b
jectiv
e
is
to
m
atch
a
s
et
o
f
ca
n
d
i
d
ate
r
esu
m
es
to
a
jo
b
d
escr
ip
tio
n
b
ased
o
n
p
r
ed
ef
in
ed
ev
alu
atio
n
cr
iter
ia
wh
ile
m
in
im
izin
g
f
als
e
p
o
s
itiv
es a
n
d
f
alse n
eg
ativ
es
.
L
et
,
,
be
=
{
1
,
2
,
.
.
.
,
}
b
e
th
e
s
et
o
f
ca
n
d
id
ate
r
esu
m
es
;
j
b
e
a
jo
b
d
escr
ip
tio
n
with
r
eq
u
ir
em
e
n
ts
=
{
1
,
2
,
.
.
.
,
}
;
=
{
1
,
2
,
.
.
.
,
}
r
ep
r
esen
t
th
e
s
et
o
f
ev
alu
atio
n
cr
iter
ia
s
u
ch
as
s
k
ills
,
ex
p
er
ien
ce
,
an
d
q
u
alif
icatio
n
s
.
T
h
e
g
o
al
is
to
lear
n
a
r
an
k
in
g
f
u
n
ctio
n
:
×
→
s
u
ch
th
at
f
o
r
ea
ch
r
esu
m
e
,
th
e
f
u
n
ctio
n
(
,
)
co
m
p
u
t
es
a
r
elev
an
ce
s
co
r
e
b
ased
o
n
tex
tu
al
s
im
ilar
ity
,
co
n
tex
tu
al
a
lig
n
m
en
t,
an
d
d
o
m
ain
-
s
p
ec
if
ic
r
eq
u
ir
em
en
ts
.
T
h
e
o
p
tim
izatio
n
o
b
jectiv
e
is
to
m
ax
im
ize
ca
n
d
id
ate
-
j
o
b
r
elev
an
ce
wh
ile
m
itig
atin
g
m
is
class
if
icatio
n
er
r
o
r
s
.
2
.
2
.
Da
t
a
prepa
ra
t
i
o
n a
nd
s
y
s
t
em
a
rc
hite
ct
ure
T
h
is
s
ec
tio
n
d
escr
ib
es
th
e
d
at
a
p
r
ep
ar
atio
n
an
d
s
y
s
tem
ar
ch
itectu
r
e
im
p
lem
en
ted
f
o
r
th
e
p
r
o
p
o
s
ed
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
s
y
s
tem
u
s
in
g
th
e
L
L
aM
A
3
m
o
d
el.
T
h
e
p
r
o
ce
s
s
b
eg
in
s
with
co
n
v
er
tin
g
u
n
s
tr
u
ctu
r
ed
r
esu
m
e
d
o
cu
m
e
n
ts
,
ty
p
ically
in
PDF
f
o
r
m
at,
in
to
s
tr
u
ctu
r
e
d
p
lain
tex
t
to
en
ab
le
d
o
w
n
s
tr
ea
m
p
r
o
ce
s
s
in
g
s
u
ch
as
em
b
ed
d
in
g
g
en
er
atio
n
,
class
if
icatio
n
,
an
d
ca
n
d
id
ate
r
an
k
i
n
g
.
T
o
p
e
r
f
o
r
m
th
is
co
n
v
er
s
io
n
,
L
M
Stu
d
io
wh
ich
is
a
s
p
ec
ialized
s
o
f
twar
e
p
latf
o
r
m
f
o
r
r
u
n
n
in
g
LLM
s
.
L
M
Stu
d
io
s
u
p
p
o
r
t
s
lo
ca
l d
ep
lo
y
m
e
n
t
o
f
th
e
L
L
aM
A
3
m
o
d
el,
e
n
ab
lin
g
ef
f
icien
t
a
n
d
p
r
iv
ate
e
x
ec
u
tio
n
with
o
u
t
r
elian
ce
o
n
clo
u
d
-
b
ased
API
s
.
T
h
e
p
latf
o
r
m
s
tr
ea
m
lin
es
th
e
tr
an
s
f
o
r
m
atio
n
o
f
r
esu
m
es
in
to
an
aly
za
b
le
tex
t
b
y
f
ac
ilit
atin
g
p
r
ec
is
e
s
eg
m
en
tatio
n
o
f
k
ey
s
ec
tio
n
s
s
u
ch
as e
d
u
ca
t
io
n
,
ex
p
er
ien
ce
,
a
n
d
s
k
ills
.
2
.
2
.
1
.
Resum
e
prepro
ce
s
s
ing
R
esu
m
es
ar
e
ty
p
ically
r
ec
eiv
ed
in
PDF
f
o
r
m
at,
wh
ich
is
u
n
s
tr
u
ctu
r
ed
an
d
u
n
s
u
itab
le
f
o
r
d
ir
ec
t
m
ac
h
in
e
p
r
o
ce
s
s
in
g
.
A
p
r
ep
r
o
ce
s
s
in
g
s
tep
is
r
eq
u
ir
e
d
to
c
o
n
v
er
t
t
h
ese
r
esu
m
es
in
to
m
a
ch
in
e
-
r
ea
d
a
b
le
tex
t.
T
h
is
was
ac
co
m
p
lis
h
ed
u
s
in
g
a
Py
t
h
o
n
s
cr
ip
t
d
ev
elo
p
e
d
with
th
e
Py
PDF2
lib
r
ar
y
.
T
h
e
s
cr
ip
t
d
ef
in
es
a
f
u
n
ctio
n
th
at
ac
ce
p
ts
th
e
f
ile
p
ath
o
f
th
e
i
n
p
u
t
PDF
an
d
th
e
d
esire
d
o
u
tp
u
t
tex
t
f
ile.
I
t
o
p
e
n
s
th
e
PDF
in
b
in
ar
y
m
o
d
e,
u
s
es
th
e
Pd
f
R
ea
d
er
cla
s
s
to
p
ar
s
e
its
co
n
ten
ts
,
an
d
ex
tr
ac
ts
tex
t
p
ag
e
b
y
p
ag
e
u
s
in
g
th
e
.
ex
tr
ac
t_
te
x
t(
)
m
eth
o
d
.
T
h
e
r
esu
ltin
g
tex
t
is
th
en
wr
itten
in
t
o
a
.
tx
t
f
ile
th
at
p
r
eser
v
es
th
e
o
r
ig
in
al
lo
g
ic
al
f
lo
w
o
f
r
esu
m
e
co
m
p
o
n
en
ts
.
T
h
ese
s
tr
u
ctu
r
ed
tex
t
f
iles
s
er
v
e
as
th
e
p
r
im
ar
y
in
p
u
t
to
th
e
L
L
aM
A
-
b
ased
p
r
o
ce
s
s
in
g
p
ip
elin
e,
en
s
u
r
in
g
th
at
r
elev
an
t
in
f
o
r
m
atio
n
s
u
ch
as
q
u
alif
icatio
n
s
,
s
k
ills
,
an
d
wo
r
k
ex
p
er
ien
ce
ca
n
b
e
ef
f
ec
tiv
el
y
in
ter
p
r
eted
b
y
th
e
la
n
g
u
a
g
e
m
o
d
el.
2
.
2
.
2
.
Sy
s
t
em
a
rc
hite
ct
ure
o
v
er
v
iew
Fig
u
r
e
1
illu
s
tr
ates
th
e
s
y
s
te
m
ar
ch
itectu
r
e
d
ev
elo
p
ed
f
o
r
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
u
s
in
g
th
e
L
L
aM
A
3
m
o
d
el.
On
ce
r
esu
m
es
h
av
e
b
ee
n
co
n
v
er
ted
to
s
tr
u
ctu
r
ed
tex
t a
n
d
p
air
ed
with
th
e
co
r
r
esp
o
n
d
i
n
g
jo
b
d
escr
ip
tio
n
,
b
o
th
in
p
u
ts
ar
e
s
im
u
ltan
eo
u
s
ly
p
r
o
ce
s
s
ed
th
r
o
u
g
h
th
e
L
L
aM
A
3
m
o
d
el
in
a
m
u
lti
-
s
tag
e
p
ip
elin
e.
I
n
itially
,
th
e
in
p
u
ts
a
r
e
tr
an
s
f
o
r
m
ed
in
to
c
o
n
tex
tu
al
em
b
e
d
d
in
g
s
th
at
ca
p
tu
r
e
s
em
an
tic
r
e
latio
n
s
h
ip
s
b
etwe
en
ca
n
d
id
ate
p
r
o
f
iles
an
d
jo
b
r
e
q
u
ir
em
en
ts
.
T
h
ese
em
b
ed
d
in
g
s
ar
e
th
en
s
u
b
jecte
d
to
r
o
o
t
m
ea
n
s
q
u
ar
e
(
R
MS)
n
o
r
m
aliza
tio
n
t
o
en
h
a
n
ce
s
tab
ilit
y
d
u
r
in
g
in
f
e
r
en
ce
.
Ne
x
t,
th
e
m
o
d
el
ap
p
lies
g
r
o
u
p
e
d
m
u
lti
-
q
u
er
y
at
ten
tio
n
with
k
ey
-
v
alu
e
ca
c
h
in
g
,
an
o
p
tim
izatio
n
tech
n
iq
u
e
th
at
allo
ws
th
e
m
o
d
el
to
ef
f
icien
tly
f
o
cu
s
o
n
th
e
m
o
s
t
r
elev
an
t
p
a
r
ts
o
f
th
e
in
p
u
t
a
cr
o
s
s
m
u
ltip
le
r
esu
m
es.
Fo
ll
o
win
g
atten
tio
n
,
th
e
o
u
tp
u
t
i
s
p
ass
ed
th
r
o
u
g
h
a
f
ee
d
f
o
r
war
d
n
eu
r
al
n
etwo
r
k
u
t
ilizin
g
a
s
witch
g
ated
lin
ea
r
u
n
it
(
SwiGLU
)
ac
tiv
atio
n
f
u
n
ct
io
n
,
wh
ic
h
h
elp
s
in
m
o
d
elin
g
c
o
m
p
lex
h
ier
ar
ch
ic
al
r
elatio
n
s
h
ip
s
.
T
h
is
o
u
tp
u
t
is
co
m
b
in
ed
with
th
e
o
r
ig
in
a
l
em
b
ed
d
in
g
s
an
d
n
o
r
m
alize
d
a
g
ain
.
Fin
ally
,
a
l
in
ea
r
tr
an
s
f
o
r
m
atio
n
is
a
p
p
lie
d
,
an
d
th
e
r
esu
ltin
g
v
ec
t
o
r
s
ar
e
p
ass
ed
th
r
o
u
g
h
a
s
o
f
tm
ax
lay
er
to
co
m
p
u
te
r
el
ev
an
ce
s
co
r
es.
T
h
ese
s
co
r
es
r
ep
r
esen
t
th
e
d
eg
r
ee
o
f
alig
n
m
en
t
b
etwe
en
ea
c
h
r
esu
m
e
an
d
th
e
jo
b
d
escr
ip
tio
n
,
en
ab
lin
g
th
e
s
y
s
tem
to
r
an
k
ca
n
d
id
ates
ac
co
r
d
in
g
ly
.
T
h
is
ar
ch
itectu
r
e
en
ab
les
h
ig
h
-
r
eso
lu
tio
n
s
em
an
tic
m
atc
h
in
g
an
d
f
o
r
m
s
th
e
b
ac
k
b
o
n
e
o
f
th
e
L
L
aM
A
-
b
ased
r
an
k
in
g
f
u
n
ctio
n
.
2
.
2
.
3
.
M
o
del deplo
y
m
ent
v
ia
L
M
St
ud
io
T
h
e
L
L
aM
A
3
m
o
d
el
em
p
lo
y
ed
in
th
is
s
tu
d
y
is
th
e
Me
ta
-
L
L
aM
A
-
3
-
8B
-
I
n
s
tr
u
ct
v
ar
ia
n
t,
ch
o
s
en
f
o
r
its
o
p
tim
al
b
alan
ce
b
etwe
en
m
o
d
el
co
m
p
lex
ity
an
d
co
m
p
u
tatio
n
al
ef
f
icien
cy
.
Dep
lo
y
m
en
t
was
ca
r
r
ied
o
u
t
u
s
in
g
L
M
Stu
d
io
,
an
o
p
en
-
s
o
u
r
ce
p
latf
o
r
m
th
at
s
u
p
p
o
r
t
s
lo
ca
l
ex
ec
u
tio
n
o
f
L
L
Ms
with
o
u
t
r
eq
u
ir
in
g
clo
u
d
-
b
ased
in
f
r
astru
ct
u
r
e.
T
h
e
m
o
d
el
was
co
n
f
i
g
u
r
ed
with
a
q
u
an
tizatio
n
lev
el
o
f
Q
4
_
K_
M,
a
co
n
tex
t
len
g
th
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
953
-
9
6
1
956
o
f
8
,
1
9
2
,
a
n
d
an
e
m
b
ed
d
in
g
s
ize
o
f
4
,
0
9
6
,
e
n
ab
lin
g
it
to
p
r
o
ce
s
s
lo
n
g
an
d
in
f
o
r
m
atio
n
-
r
i
ch
in
p
u
t
s
eq
u
e
n
ce
s
ef
f
ec
tiv
ely
.
Ar
ch
itectu
r
ally
,
it
in
clu
d
es
3
2
tr
an
s
f
o
r
m
er
lay
er
s
an
d
3
2
atten
tio
n
h
ea
d
s
,
with
8
h
ea
d
s
d
esig
n
ated
f
o
r
k
ey
-
v
alu
e
ca
ch
i
n
g
,
en
h
an
cin
g
atten
tio
n
p
er
f
o
r
m
a
n
ce
d
u
r
in
g
in
f
e
r
en
ce
.
Po
s
itio
n
al
en
co
d
in
g
was
h
an
d
led
u
s
in
g
r
o
tar
y
p
o
s
itio
n
em
b
ed
d
in
g
(
R
o
PE)
with
a
f
r
e
q
u
en
c
y
b
ase
o
f
5
0
0
,
0
0
0
,
wh
ich
im
p
r
o
v
es
th
e
m
o
d
el’
s
ab
ilit
y
to
u
n
d
er
s
tan
d
to
k
en
o
r
d
er
in
lo
n
g
s
eq
u
en
ce
s
.
T
o
f
in
e
-
tu
n
e
th
e
g
en
e
r
atio
n
q
u
ality
an
d
d
iv
e
r
s
ity
,
tem
p
er
at
u
r
e
was
s
et
to
0
.
8
,
to
p
-
k
s
am
p
lin
g
to
4
0
,
an
d
a
r
ep
ea
t
p
en
alty
o
f
1
.
1
was
ap
p
lied
to
d
is
co
u
r
ag
e
r
ep
etitiv
e
o
u
tp
u
ts
.
Fo
r
im
p
r
o
v
e
d
ef
f
icien
cy
,
GPU
ac
ce
ler
atio
n
u
s
in
g
NVI
DI
A
C
UDA
was
en
ab
led
,
an
d
C
PU
m
u
ltit
h
r
ea
d
in
g
was
o
p
tim
ized
to
r
ed
u
ce
laten
cy
.
T
h
e
f
u
ll
m
o
d
el
was
lo
ad
ed
in
to
R
AM
to
s
u
p
p
o
r
t
f
ast
in
f
er
en
ce
,
an
d
p
ar
a
m
eter
s
s
u
ch
as
p
r
o
m
p
t
ev
alu
atio
n
b
atch
s
ize
an
d
c
o
n
tex
t
win
d
o
w
wer
e
ad
ju
s
ted
b
ased
o
n
s
y
s
tem
ca
p
ab
ilit
ies
an
d
d
ata
s
ize.
T
h
is
co
n
f
ig
u
r
atio
n
en
s
u
r
ed
s
m
o
o
th
an
d
ef
f
icien
t
ex
ec
u
tio
n
o
f
t
h
e
r
esu
m
e
s
cr
ee
n
in
g
wo
r
k
f
lo
w
wh
ile
m
ain
tain
in
g
h
ig
h
ac
cu
r
ac
y
in
ca
n
d
id
ate
-
jo
b
m
atch
in
g
.
Fig
u
r
e
2
illu
s
tr
ates
th
e
co
m
p
lete
co
n
f
ig
u
r
atio
n
s
etu
p
u
s
ed
in
th
is
s
tu
d
y
.
Sp
ec
if
ically
,
Fig
u
r
e
2
(
a)
p
r
es
en
ts
th
e
m
o
d
el
s
elec
tio
n
an
d
in
itializatio
n
p
ar
am
eter
s
,
in
clu
d
in
g
th
e
s
elec
ted
L
L
aM
A
3
v
ar
ian
t
an
d
p
r
o
m
p
t
f
o
r
m
attin
g
o
p
tio
n
s
tailo
r
ed
f
o
r
r
esu
m
e
s
cr
ee
n
in
g
task
s
.
Fig
u
r
e
2
(
b
)
s
h
o
ws
th
e
h
ar
d
war
e
r
eso
u
r
ce
allo
ca
tio
n
,
h
ig
h
lig
h
tin
g
GPU
co
r
e
u
tili
za
tio
n
an
d
m
em
o
r
y
o
f
f
lo
a
d
in
g
s
ettin
g
s
ad
o
p
ted
to
en
s
u
r
e
ef
f
icien
t
in
f
er
e
n
ce
.
Fig
u
r
e
2
(
c
)
d
ep
icts
th
e
in
f
er
en
c
e
an
d
h
ar
d
war
e
p
ar
am
eter
s
,
s
u
ch
as
tem
p
er
atu
r
e
,
to
k
en
g
e
n
er
atio
n
lim
its
,
an
d
s
am
p
lin
g
s
tr
ateg
ies,
wh
ich
c
o
l
lectiv
ely
in
f
lu
en
ce
th
e
m
o
d
el
’
s
r
esp
o
n
s
e
q
u
ality
an
d
co
n
s
is
ten
cy
.
Fig
u
r
e
1
.
Pro
p
o
s
ed
s
y
s
tem
ar
c
h
itectu
r
e
f
o
r
au
to
m
ate
d
r
esu
m
e
s
cr
ee
n
in
g
u
s
in
g
L
L
aM
A
3
m
o
d
el
(
a)
(
b
)
(
c)
Fig
u
r
e
2
.
C
o
n
f
ig
u
r
atio
n
s
ettin
g
s
o
f
L
L
aM
A
3
f
o
r
a
u
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
of
(
a)
m
o
d
e
l s
elec
tio
n
in
LM
Stu
d
io
,
(
b
)
GPU
co
r
e
co
u
n
t a
n
d
m
e
m
o
r
y
allo
ca
tio
n
s
ettin
g
s
f
o
r
L
L
aM
A
3
m
o
d
el
d
ep
l
o
y
m
en
t
,
a
n
d
(
c)
h
ar
d
war
e
p
ar
a
m
eter
s
s
etu
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ca
la
b
le
r
esu
me
s
creen
in
g
u
s
i
n
g
la
r
g
e
la
n
g
u
a
g
e
mo
d
el
Meta
A
I
ve
r
s
io
n
3
(
A
s
mita
Desh
mu
kh
)
957
2
.
2
.
4
.
P
ro
m
pt
eng
ineering
a
nd
ex
ec
utio
n
T
o
in
ter
ac
t
with
th
e
m
o
d
el,
s
p
ec
if
ic
p
r
o
m
p
ts
wer
e
cr
af
ted
f
o
r
b
o
t
h
th
e
jo
b
d
escr
ip
tio
n
a
n
d
ca
n
d
id
ate
r
esu
m
es.
T
h
e
jo
b
d
escr
ip
tio
n
was
in
p
u
t
u
s
in
g
th
e
co
m
m
an
d
:
“I
n
ee
d
y
o
u
r
h
el
p
in
d
ec
id
in
g
am
o
n
g
all
ca
n
d
id
ates w
h
ich
o
n
e
is
th
e
b
e
s
t f
it f
o
r
th
is
jo
b
d
escr
ip
tio
n
: [
J
o
b
Descr
ip
tio
n
]
.
”
E
ac
h
r
esu
m
e
was c
o
n
v
er
ted
to
s
tr
u
ctu
r
ed
tex
t
an
d
co
n
ca
ten
ated
in
to
a
s
in
g
le
in
p
u
t
f
ile.
T
h
e
f
o
llo
win
g
p
r
o
m
p
t
was
u
s
ed
f
o
r
s
cr
ee
n
in
g
:
“Bas
ed
o
n
th
e
jo
b
d
escr
ip
tio
n
an
d
th
e
p
r
o
v
id
ed
ca
n
d
id
ate
p
r
o
f
iles
,
h
elp
m
e
d
ec
id
e
w
h
ich
c
an
d
id
ate
is
th
e
b
est
f
it
f
o
r
t
h
e
jo
b
u
s
in
g
th
e
r
esu
m
e
s
u
m
m
ar
y
o
f
ea
c
h
c
an
d
id
ate'
s
s
k
ills
,
ed
u
ca
tio
n
,
ac
h
iev
em
en
ts
,
a
n
d
ex
p
er
ien
ce
.
”
L
L
aM
A
3
p
r
o
ce
s
s
ed
th
ese
in
p
u
ts
an
d
d
ir
ec
tly
r
etu
r
n
ed
th
e
n
am
e
o
f
th
e
m
o
s
t
s
u
itab
le
ca
n
d
id
ate,
alo
n
g
with
a
r
ea
s
o
n
ed
e
x
p
lan
a
tio
n
f
o
r
th
e
s
elec
tio
n
.
3.
RE
SU
L
T
S
Fo
r
th
is
s
tu
d
y
,
th
e
L
L
aM
A
3
m
o
d
el
(
8
B
v
er
s
io
n
)
was
d
e
p
lo
y
ed
in
L
M
Stu
d
io
,
r
eq
u
ir
in
g
at
least
8
GB
GPU
an
d
more
t
h
an
8
GB
R
AM
.
Key
h
y
p
er
p
ar
am
et
er
s
s
u
ch
as
tem
p
er
atu
r
e
(
0
.
8
)
an
d
to
p
-
k
(
4
0
)
wer
e
f
in
e
-
tu
n
e
d
to
b
alan
ce
d
iv
er
s
ity
an
d
d
eter
m
i
n
is
m
.
R
esu
m
e
PD
Fs
wer
e
co
n
v
er
ted
to
tex
t u
s
in
g
Py
PDF2
,
an
d
all
r
esu
m
es we
r
e
ev
alu
ated
u
s
in
g
th
e
s
am
e
s
tan
d
ar
d
ized
p
r
o
m
p
t.
3
.
1
.
E
x
ec
utio
n t
i
m
e
a
nd
ba
t
ch
a
na
ly
s
is
T
o
ev
alu
ate
th
e
s
ca
lab
ilit
y
an
d
p
r
o
ce
s
s
in
g
ef
f
icien
cy
o
f
t
h
e
p
r
o
p
o
s
ed
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
s
y
s
tem
,
ex
p
er
im
e
n
ts
wer
e
c
o
n
d
u
cted
b
y
v
a
r
y
in
g
th
e
n
u
m
b
er
o
f
r
esu
m
es
p
r
o
ce
s
s
ed
p
e
r
b
atch
.
Fig
u
r
e
3
illu
s
tr
ates
th
e
s
y
s
tem
p
er
f
o
r
m
an
ce
in
ter
m
s
o
f
ex
ec
u
tio
n
ti
m
e
an
d
tex
t
v
o
lu
m
e
p
r
o
ce
s
s
ed
.
Fig
u
r
e
3
(
a)
s
h
o
ws
th
at
a
b
atch
o
f
th
r
ee
r
esu
m
es
to
o
k
ap
p
r
o
x
im
ately
2
s
ec
o
n
d
s
to
p
r
o
ce
s
s
,
av
er
ag
in
g
0
.
6
6
s
ec
o
n
d
s
p
er
r
esu
m
e.
As
b
atch
s
ize
in
cr
ea
s
es,
th
e
to
tal
ex
ec
u
tio
n
tim
e
s
h
o
ws
a
n
o
n
-
lin
ea
r
tr
en
d
,
lik
ely
d
u
e
to
GPU
m
em
o
r
y
lim
itatio
n
s
.
Fig
u
r
e
3
(
b
)
d
is
p
l
ay
s
a
s
lig
h
t
d
ec
r
ea
s
e
in
th
e
a
v
er
ag
e
w
o
r
d
co
u
n
t
p
e
r
r
esu
m
e
as
th
e
b
atch
s
ize
in
cr
ea
s
es,
s
u
g
g
esti
n
g
p
o
s
s
ib
le
to
k
en
tr
u
n
ca
tio
n
o
r
co
n
tex
t
-
le
n
g
th
p
r
io
r
itizatio
n
.
(
a)
(
b
)
Fig
u
r
e
3
.
Per
f
o
r
m
an
c
e
an
aly
s
i
s
o
f
th
e
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
s
y
s
tem
of
(
a)
n
u
m
b
e
r
o
f
r
esu
m
es v
er
s
u
s
to
tal
ex
ec
u
tio
n
tim
e
a
n
d
(
b
)
av
er
ag
e
n
u
m
b
er
o
f
wo
r
d
s
p
r
o
ce
s
s
ed
p
er
r
esu
m
e
i
n
a
b
atch
An
aly
s
is
in
d
icate
s
th
at
p
r
o
ce
s
s
in
g
a
b
atch
o
f
th
r
ee
r
esu
m
es
tak
es
ap
p
r
o
x
im
ately
2
s
ec
o
n
d
s
,
r
esu
ltin
g
in
an
av
er
ag
e
s
ca
n
n
in
g
s
p
ee
d
o
f
0
.
6
6
s
ec
o
n
d
s
p
er
r
es
u
m
e.
As
th
e
n
u
m
b
er
o
f
r
esu
m
es
in
cr
ea
s
es,
th
e
p
r
o
ce
s
s
in
g
tim
e
is
ex
p
ec
ted
t
o
f
o
llo
w
a
n
o
n
-
lin
ea
r
tr
en
d
a
n
d
m
ay
e
v
en
tu
ally
p
latea
u
d
u
e
to
p
o
ten
tial
GPU
m
em
o
r
y
lim
itatio
n
s
.
T
o
im
p
r
o
v
e
p
r
o
ce
s
s
in
g
e
f
f
icien
cy
,
b
a
tch
p
r
o
ce
s
s
in
g
s
tr
ateg
ies
s
im
ilar
to
th
e
b
u
b
b
le
s
o
r
tin
g
alg
o
r
ith
m
ca
n
b
e
em
p
lo
y
ed
.
T
h
e
p
l
o
t
in
Fig
u
r
e
3
(
b
)
d
is
p
lay
s
th
e
av
er
ag
e
wo
r
d
co
u
n
t
o
f
r
esu
m
es
p
r
o
ce
s
s
ed
b
y
th
e
L
L
aM
A
3
m
o
d
el
f
o
r
b
atch
s
izes
r
an
g
in
g
f
r
o
m
1
to
3
.
T
h
e
x
-
a
x
is
r
ep
r
e
s
en
ts
th
e
n
u
m
b
er
o
f
r
esu
m
es
in
a
b
atch
,
wh
ile
th
e
y
-
ax
is
s
h
o
ws
th
e
av
er
ag
e
n
u
m
b
er
o
f
wo
r
d
s
p
er
r
esu
m
e.
T
h
e
d
ata
s
u
g
g
ests
a
s
lig
h
t
d
ec
r
ea
s
e
in
av
er
ag
e
wo
r
d
co
u
n
t
p
er
r
esu
m
e
as
th
e
b
atch
s
ize
in
cr
ea
s
es,
f
o
r
ex
am
p
le
with
3
3
6
wo
r
d
s
f
o
r
a
s
in
g
le
r
esu
m
e
an
d
2
0
2
wo
r
d
s
f
o
r
a
b
atch
o
f
t
h
r
ee
.
3
.
2
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
a
nd
f
ina
l dec
is
io
n o
utput
F
i
g
u
r
e
4
d
i
s
p
l
a
y
s
a
co
m
p
r
eh
en
s
i
v
e
co
m
p
a
r
a
t
iv
e
an
a
l
y
s
i
s
t
ab
l
e
g
en
e
r
a
t
e
d
b
y
L
L
a
M
A
3
,
ev
a
l
u
a
t
in
g
t
h
r
e
e
c
a
n
d
i
d
a
te
s
(
A
b
h
i
n
a
v
D
e
o
t
a
l
e
,
H
a
r
i
t
P
a
t
w
a,
an
d
A
b
h
i
s
h
ek
W
a
l
a
w
a
lk
a
r
)
a
g
a
in
s
t
t
h
e
s
p
e
c
i
f
i
ed
jo
b
r
e
q
u
i
r
em
e
n
t
s
.
T
h
e
t
ab
u
la
r
o
u
t
p
u
t
d
e
m
o
n
s
tr
a
t
e
s
t
h
e
m
o
d
e
l
's
c
a
p
a
b
i
l
i
t
y
to
s
y
s
t
e
m
a
t
i
ca
l
l
y
a
n
a
l
y
z
e
a
n
d
c
a
t
e
g
o
r
i
z
e
c
an
d
id
a
t
e
q
u
a
l
i
f
ica
t
i
o
n
s
a
c
r
o
s
s
m
u
l
t
i
p
le
d
i
m
en
s
i
o
n
s
l
i
k
e,
s
k
i
l
l
s
a
s
s
e
s
s
m
en
t
h
e
r
e
th
e
o
u
tp
u
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
953
-
9
6
1
958
p
r
e
s
e
n
t
s
a
d
e
t
a
i
le
d
b
r
e
ak
d
o
wn
o
f
e
a
ch
c
a
n
d
i
d
a
t
e
's
t
e
c
h
n
i
ca
l
a
n
d
s
o
f
t
s
k
i
l
l
s
,
h
i
g
h
l
ig
h
t
in
g
t
h
e
ir
r
e
l
ev
an
c
e
to
t
h
e
p
o
s
i
t
io
n
r
eq
u
ir
e
m
en
t
s
.
Ne
x
t
i
s
t
h
e
ex
p
er
i
en
c
e
e
v
a
lu
at
i
o
n
w
h
er
e
t
h
e
m
o
d
e
l
q
u
an
t
if
i
e
s
an
d
q
u
a
l
i
f
ie
s
p
r
o
f
e
s
s
i
o
n
a
l
ex
p
er
i
e
n
c
e
,
p
r
o
v
i
d
i
n
g
n
o
t
ju
s
t
d
u
r
a
t
i
o
n
m
e
tr
i
cs
b
u
t
a
l
s
o
co
n
te
x
tu
a
l
an
a
ly
s
i
s
o
f
r
o
l
e
r
e
l
e
v
a
n
c
e
t
o
t
h
e
t
a
r
g
e
t
p
o
s
i
t
i
o
n
.
T
h
en
s
t
a
n
d
s
th
e
ed
u
c
a
t
io
n
a
l
q
u
a
l
i
f
ica
t
i
o
n
a
n
a
ly
s
i
s
b
ey
o
n
d
l
i
s
t
i
n
g
d
e
g
r
e
e
s
,
th
e
t
ab
l
e
i
n
c
l
u
d
e
s
an
e
v
a
l
u
a
t
io
n
o
f
e
d
u
c
a
t
i
o
n
a
l
r
e
l
ev
a
n
ce
to
th
e
j
o
b
r
e
q
u
i
r
em
e
n
t
s
,
d
e
m
o
n
s
t
r
a
t
in
g
L
L
a
M
A
3
's
a
b
i
l
i
ty
t
o
co
n
te
x
tu
a
l
i
z
e
ed
u
c
a
t
io
n
a
l
b
ac
k
g
r
o
u
n
d
s
.
F
u
r
th
er
i
s
t
h
e
a
ch
i
e
v
em
e
n
t
r
e
c
o
g
n
i
t
io
n
b
ei
n
g
i
n
t
h
e
n
o
ta
b
l
e
a
c
c
o
m
p
l
i
s
h
m
e
n
t
s
a
r
e
e
x
tr
a
c
te
d
an
d
p
r
e
s
e
n
te
d
w
i
t
h
a
n
im
p
l
i
c
i
t
w
e
i
g
h
t
i
n
g
to
w
a
r
d
ac
h
i
e
v
e
m
en
t
s
m
o
s
t
r
e
l
e
v
an
t
t
o
t
h
e
p
o
s
i
t
i
o
n
an
d
w
i
t
h
t
h
e
l
a
s
t
b
e
i
n
g
d
e
c
i
s
i
o
n
s
u
p
p
o
r
t
v
i
s
u
a
l
i
z
a
t
io
n
wh
i
ch
w
i
t
h
i
t
s
s
t
r
u
c
tu
r
ed
f
o
r
m
a
t
en
a
b
l
e
s
r
a
p
id
i
d
en
t
i
f
i
ca
t
i
o
n
o
f
s
t
r
en
g
th
s
a
n
d
g
ap
s
a
cr
o
s
s
c
a
n
d
i
d
a
te
s
,
s
e
r
v
in
g
a
s
a
n
e
f
f
e
c
t
iv
e
d
e
c
i
s
i
o
n
s
u
p
p
o
r
t
to
o
l
.
T
h
i
s
o
u
t
p
u
t
r
ep
r
e
s
e
n
t
s
a
s
i
g
n
i
f
i
ca
n
t
ad
v
an
c
e
m
e
n
t
o
v
er
t
r
ad
i
t
i
o
n
a
l
k
e
y
w
o
r
d
-
b
a
s
ed
r
e
s
u
m
e
s
c
r
e
e
n
in
g
b
y
p
r
o
v
i
d
in
g
s
t
an
d
ar
d
i
z
ed
,
m
u
l
t
i
-
f
a
c
e
t
ed
c
an
d
i
d
a
t
e
co
m
p
a
r
i
s
o
n
s
.
T
h
e
e
x
e
cu
t
i
o
n
t
i
m
e
o
f
2
.
2
9
s
e
co
n
d
s
f
o
r
g
en
er
a
t
i
n
g
t
h
i
s
c
o
m
p
r
e
h
en
s
i
v
e
an
a
ly
s
i
s
f
u
r
th
e
r
d
e
m
o
n
s
tr
a
t
e
s
t
h
e
ef
f
i
c
i
en
c
y
g
a
in
o
v
e
r
m
a
n
u
a
l
r
e
v
ie
w
p
r
o
c
e
s
s
e
s
,
wh
i
c
h
t
y
p
i
c
a
l
ly
r
e
q
u
i
r
e
1
5
-
3
0
m
i
n
u
t
e
s
p
er
r
e
s
u
m
e.
H
ir
i
n
g
d
e
c
i
s
io
n
s
c
a
n
b
e
a
f
f
e
c
te
d
b
y
b
ia
s
f
r
o
m
i
n
i
t
ia
l
in
f
o
r
m
a
t
io
n
,
l
e
a
d
in
g
r
e
cr
u
i
t
er
s
t
o
f
a
v
o
r
c
a
n
d
i
d
a
te
s
b
a
s
e
d
o
n
e
a
r
l
y
im
p
r
e
s
s
i
o
n
s
r
a
t
h
er
th
a
n
o
v
e
r
a
l
l
q
u
a
l
i
f
i
ca
t
io
n
s
.
Fig
u
r
e
5
,
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
o
u
tp
u
t
g
en
er
ated
b
y
L
L
aM
A
3
,
wh
ich
id
en
tifie
s
th
e
ca
n
d
id
ate
d
ee
m
ed
to
b
e
th
e
b
est
f
it
f
o
r
th
e
jo
b
r
o
le.
Ad
d
itio
n
ally
,
th
e
o
u
tp
u
t
p
r
o
v
id
es
a
d
etaile
d
r
atio
n
ale
b
eh
i
n
d
L
L
aM
A
3
’
s
d
ec
is
io
n
,
o
f
f
e
r
in
g
in
s
ig
h
ts
in
to
t
h
e
f
ac
to
r
s
th
at
m
ak
e
th
is
p
a
r
ticu
lar
ca
n
d
id
at
e
th
e
m
o
s
t
s
u
itab
le
f
o
r
th
e
p
o
s
itio
n
.
T
h
is
co
m
p
r
eh
en
s
iv
e
an
aly
s
is
n
o
t
o
n
ly
f
ac
ilit
ates
th
e
s
elec
tio
n
p
r
o
ce
s
s
b
u
t
also
en
h
an
ce
s
tr
an
s
p
ar
en
cy
b
y
p
r
o
v
i
d
in
g
cl
ea
r
ju
s
tific
atio
n
s
f
o
r
th
e
m
o
d
el’
s
d
ec
is
io
n
-
m
ak
in
g
.
L
ev
e
r
ag
in
g
L
L
aM
A
3
’
s
ad
v
an
ce
d
ca
p
ab
ilit
ies,
r
ec
r
u
it
er
s
,
an
d
h
ir
in
g
m
an
a
g
er
s
ca
n
m
ak
e
m
o
r
e
in
f
o
r
m
ed
an
d
o
b
jectiv
e
d
ec
is
io
n
s
,
u
ltima
tely
im
p
r
o
v
in
g
th
e
q
u
al
ity
an
d
ef
f
icien
cy
o
f
th
e
r
ec
r
u
itm
en
t
p
r
o
ce
s
s
.
I
t
is
also
im
p
o
r
tan
t
to
n
o
tin
g
th
a
t
th
is
d
ata
o
n
ly
s
h
o
ws
a
v
er
y
s
m
all
s
am
p
le
s
ize,
s
o
it’s
n
o
t
p
o
s
s
ib
le
to
d
r
aw
an
y
d
ef
in
itiv
e
co
n
clu
s
io
n
s
f
r
o
m
it.
I
t is also
u
n
clea
r
wh
eth
er
t
h
is
tr
en
d
wo
u
ld
h
o
ld
f
o
r
lar
g
e
r
b
at
ch
es o
f
r
esu
m
es.
Fig
u
r
e
4
.
C
o
m
p
a
r
ativ
e
an
aly
s
i
s
o
f
th
r
ee
ca
n
d
id
ates e
v
alu
ated
in
a
s
in
g
le
b
atch
u
s
in
g
L
L
aM
A
3
Fig
u
r
e
5
.
Fin
al
ca
n
d
id
ate
s
elec
tio
n
b
ased
o
n
jo
b
d
escr
ip
tio
n
a
n
d
r
esu
m
e
a
n
aly
s
is
u
s
in
g
L
L
a
MA
3
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ca
la
b
le
r
esu
me
s
creen
in
g
u
s
i
n
g
la
r
g
e
la
n
g
u
a
g
e
mo
d
el
Meta
A
I
ve
r
s
io
n
3
(
A
s
mita
Desh
mu
kh
)
959
4.
DIS
CU
SS
I
O
N
T
h
e
p
r
o
p
o
s
ed
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
ap
p
r
o
ac
h
,
lev
er
ag
in
g
th
e
ca
p
ab
ilit
ies
o
f
th
e
L
L
aM
A
3
LLM
,
o
f
f
er
s
s
ev
er
al
d
is
tin
ct
ad
v
an
tag
es
o
v
e
r
tr
ad
itio
n
al
m
e
th
o
d
s
.
First,
it
s
ig
n
if
ican
tly
ac
ce
ler
ates
th
e
in
itial
s
cr
ee
n
in
g
p
r
o
ce
s
s
b
y
r
a
p
id
ly
an
aly
zin
g
lar
g
e
v
o
lu
m
es
o
f
r
esu
m
es,
ex
tr
ac
tin
g
r
elev
an
t
in
f
o
r
m
atio
n
,
a
n
d
ef
f
icien
tly
id
en
tif
y
in
g
t
o
p
ca
n
d
id
ates.
T
h
is
tim
e
-
s
av
in
g
asp
e
ct
allo
ws
r
ec
r
u
iter
s
an
d
h
ir
in
g
m
an
ag
er
s
to
f
o
c
u
s
th
eir
ef
f
o
r
ts
o
n
in
-
d
e
p
th
e
v
al
u
atio
n
s
o
f
th
e
m
o
s
t
p
r
o
m
is
in
g
ap
p
lican
ts
.
M
o
r
eo
v
er
,
L
L
a
MA
3
’
s
co
n
te
x
tu
al
u
n
d
er
s
tan
d
i
n
g
o
f
lan
g
u
ag
e
en
ab
les
it
to
m
o
v
e
b
ey
o
n
d
s
im
p
lis
tic
k
ey
wo
r
d
m
atch
in
g
,
en
h
a
n
cin
g
th
e
ac
cu
r
ac
y
o
f
ca
n
d
id
ate
e
v
alu
atio
n
s
.
B
y
h
o
lis
tically
an
aly
zin
g
r
esu
m
es,
th
e
m
o
d
el
ca
n
id
en
tif
y
q
u
alif
i
ed
in
d
i
v
id
u
als
wh
o
m
ay
n
o
t
u
s
e
ex
ac
t
k
e
y
wo
r
d
p
h
r
ases
b
u
t
p
o
s
s
ess
th
e
r
elev
an
t
s
k
ills
an
d
ex
p
er
ien
ce
r
eq
u
ir
e
d
f
o
r
th
e
r
o
le.
T
h
e
ap
p
licatio
n
s
o
f
th
is
AI
-
d
r
i
v
en
ap
p
r
o
ac
h
e
x
ten
d
b
ey
o
n
d
tr
a
d
itio
n
al
jo
b
r
ec
r
u
itm
e
n
t.
I
t
c
o
u
ld
b
e
ad
ap
ted
f
o
r
in
ter
n
al
em
p
lo
y
ee
p
r
o
m
o
tio
n
p
r
o
ce
s
s
es,
s
tr
ea
m
lin
in
g
th
e
id
en
tific
atio
n
o
f
s
u
itab
le
ca
n
d
id
ates
with
in
an
o
r
g
an
izatio
n
b
ased
o
n
th
eir
s
k
ills
an
d
ex
p
er
ien
ce
.
Ad
d
itio
n
ally
,
ca
r
ee
r
co
u
n
s
ellin
g
s
er
v
ices
co
u
ld
lev
er
ag
e
th
is
tech
n
o
lo
g
y
to
m
atch
in
d
iv
id
u
als
with
ap
p
r
o
p
r
iate
jo
b
o
p
p
o
r
tu
n
ities
b
ased
o
n
th
eir
q
u
alif
icatio
n
s
an
d
ca
r
ee
r
g
o
als.
An
ex
p
er
im
en
t w
as c
o
n
d
u
cted
to
co
m
p
ar
e
L
L
aM
A
3
with
o
t
h
er
lead
in
g
L
L
Ms f
o
r
r
esu
m
e
s
cr
ee
n
in
g
,
with
r
esu
lts
s
h
o
wn
in
T
ab
le
1
.
Me
tr
ics s
u
ch
as a
cc
u
r
ac
y
,
p
r
ec
is
io
n
,
an
d
r
ec
all
wer
e
u
s
ed
,
a
n
d
wh
ile
L
L
aM
A
3
s
h
o
wed
s
tr
o
n
g
r
esu
lts
,
GPT
-
4
ac
h
iev
e
d
th
e
h
ig
h
est
av
er
ag
e
s
co
r
e.
L
L
aM
A
3
,
wh
e
n
im
p
lem
en
ted
i
n
L
M
Stu
d
io
,
s
u
p
p
o
r
ts
a
m
ax
im
u
m
in
p
u
t
o
f
4
,
096
t
o
k
en
s
(
~3
,
0
0
0
wo
r
d
s
)
.
Giv
en
th
e
a
v
er
ag
e
r
esu
m
e
len
g
th
o
f
5
0
0
wo
r
d
s
,
ar
o
u
n
d
s
ix
r
esu
m
es
ca
n
b
e
p
r
o
ce
s
s
ed
p
er
in
p
u
t.
T
h
is
s
tu
d
y
in
tr
o
d
u
ce
s
a
n
o
v
el
ap
p
licatio
n
o
f
L
L
aM
A
3
f
o
r
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
,
lev
er
a
g
in
g
its
co
n
tex
tu
al
u
n
d
er
s
tan
d
in
g
an
d
o
f
f
lin
e
d
ep
l
o
y
m
e
n
t
th
r
o
u
g
h
L
M
Stu
d
io
to
a
d
d
r
es
s
p
r
iv
ac
y
co
n
ce
r
n
s
.
I
t
p
r
esen
ts
em
p
ir
ical
d
ata
o
n
p
er
f
o
r
m
an
ce
,
in
clu
d
in
g
a
n
av
er
ag
e
r
u
n
tim
e
o
f
0
.
6
6
s
ec
o
n
d
s
p
er
r
esu
m
e
in
s
m
all
b
atch
es,
an
d
h
ig
h
lig
h
ts
L
L
aM
A
3
’
s
ab
ilit
y
to
p
r
o
v
id
e
co
m
p
ar
ativ
e
ca
n
d
id
ate
an
aly
s
is
with
ex
p
lan
atio
n
s
.
GPU
m
e
m
o
r
y
lim
itatio
n
s
an
d
b
atch
p
r
o
ce
s
s
in
g
s
tr
ateg
ies
ar
e
d
is
cu
s
s
ed
,
alo
n
g
with
et
h
ical
co
n
ce
r
n
s
lik
e
p
o
ten
tial m
o
d
el
b
ias.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
co
u
l
d
b
en
ef
it
r
ec
r
u
itm
en
t
b
y
en
a
b
lin
g
m
o
r
e
ef
f
icien
t
jo
b
m
a
tch
in
g
an
d
p
o
ten
tially
r
e
d
u
cin
g
u
n
em
p
lo
y
m
en
t.
I
ts
o
b
jectiv
e
ass
ess
m
en
ts
m
ay
h
elp
m
in
im
ize
h
u
m
an
b
ias,
f
o
s
ter
in
g
in
clu
s
iv
ity
,
an
d
f
air
n
ess
.
B
y
au
to
m
atin
g
in
itial
s
cr
ee
n
in
g
,
r
ec
r
u
iter
s
ca
n
f
o
c
u
s
o
n
in
-
d
e
p
th
ev
alu
atio
n
a
n
d
cu
ltu
r
al
f
it.
T
h
e
m
o
d
el
also
h
as
p
o
ten
tial
a
p
p
licatio
n
s
i
n
ca
r
ee
r
c
o
u
n
s
ellin
g
an
d
in
ter
n
al
p
r
o
m
o
tio
n
s
.
Ho
wev
er
,
its
ef
f
ec
tiv
en
ess
d
e
p
en
d
s
o
n
d
iv
er
s
e,
u
n
b
iased
tr
a
in
in
g
d
ata.
W
h
ile
n
o
t
a
n
ew
a
lg
o
r
ith
m
,
th
is
wo
r
k
in
teg
r
ates
L
L
aM
A
3
in
to
a
co
m
p
lete
wo
r
k
f
lo
w
in
v
o
lv
in
g
P
DF
-
to
-
tex
t
co
n
v
er
s
io
n
,
jo
b
d
e
s
cr
ip
tio
n
p
r
o
m
p
tin
g
,
an
d
b
atch
o
p
tim
izatio
n
with
in
L
M
Stu
d
io
.
T
h
e
s
y
s
tem
s
h
o
u
ld
s
u
p
p
o
r
t,
n
o
t
r
ep
lace
,
h
u
m
an
ju
d
g
m
en
t.
Desp
ite
lim
itatio
n
s
,
L
L
aM
A
3
s
h
o
ws s
ig
n
if
ican
t p
r
o
m
is
e
in
m
o
d
er
n
i
zin
g
an
d
im
p
r
o
v
in
g
r
ec
r
u
itm
e
n
t p
r
ac
tices.
T
ab
le
1
.
C
o
m
p
a
r
is
o
n
o
f
r
esu
m
e
s
cr
ee
n
in
g
p
e
r
f
o
r
m
an
ce
u
s
in
g
d
if
f
er
en
t
L
L
M
s
M
o
d
e
l
A
v
e
r
a
g
e
sc
o
r
e
LLa
M
A
3
6
6
.
9
2
G
P
T
-
4
7
3
.
8
5
C
l
a
u
d
e
3
S
o
n
n
e
t
6
4
.
6
6
G
e
mi
n
i
P
r
o
1
.
5
6
9
.
1
0
5.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
h
ig
h
lig
h
ts
th
e
s
ig
n
if
ican
t
p
o
ten
tial
o
f
L
L
Ms,
p
ar
ticu
lar
ly
Me
ta
AI
’
s
L
L
aM
A
3
,
in
tr
an
s
f
o
r
m
in
g
au
to
m
ated
r
esu
m
e
s
cr
ee
n
in
g
s
y
s
tem
s
.
lev
an
t
s
k
ills
an
d
ex
p
er
ien
ce
s
an
d
m
a
tch
in
g
ca
n
d
i
d
ates
to
jo
b
d
escr
ip
tio
n
s
m
o
r
e
ac
c
u
r
ately
th
an
tr
ad
itio
n
al
k
e
y
wo
r
d
-
b
ased
m
eth
o
d
s
.
B
y
em
p
l
f
o
y
in
g
L
L
aM
A
3
’
s
ad
v
an
ce
d
n
atu
r
al
lan
g
u
a
g
e
u
n
d
er
s
tan
d
in
g
ca
p
ab
ilit
ies,
th
e
p
r
o
p
o
s
ed
s
y
s
tem
s
u
r
p
ass
es
tr
ad
itio
n
al
k
e
y
wo
r
d
-
b
ased
ap
p
r
o
ac
h
es
th
r
o
u
g
h
co
n
tex
tu
al
an
aly
s
is
o
f
ca
n
d
id
ate
p
r
o
f
iles
.
Un
lik
e
ea
r
lier
s
y
s
te
m
s
th
at
r
ely
h
ea
v
ily
o
n
s
u
r
f
ac
e
-
lev
el
ter
m
m
atch
i
n
g
,
L
L
aM
A
3
e
f
f
ec
tiv
ely
i
n
ter
p
r
ets
th
e
s
em
an
tic
r
elev
an
c
e
o
f
a
ca
n
d
i
d
ate’
s
s
k
ills
,
q
u
alif
icatio
n
s
,
an
d
ex
p
er
ien
ce
in
r
elatio
n
to
s
p
ec
if
ic
jo
b
d
escr
ip
tio
n
s
.
T
h
e
s
y
s
tem
d
em
o
n
s
tr
ated
a
n
av
er
ag
e
s
ca
n
n
in
g
s
p
ee
d
o
f
0
.
6
6
s
ec
o
n
d
s
p
er
r
esu
m
e
f
o
r
s
m
all
b
atch
es,
th
o
u
g
h
lar
g
er
b
atch
es
m
ay
r
eq
u
ir
e
o
p
tim
ized
p
r
o
ce
s
s
in
g
d
u
e
to
GPU
co
n
s
tr
ain
ts
.
L
L
aM
A
3
also
o
u
tp
er
f
o
r
m
ed
b
asic
k
ey
wo
r
d
ap
p
r
o
ac
h
es
b
y
ev
alu
atin
g
ca
n
d
id
ate
q
u
alif
ic
atio
n
s
h
o
lis
tically
.
Ho
wev
er
,
s
ca
lab
ilit
y
ch
allen
g
es
wer
e
n
o
ted
as
b
atc
h
s
ize
in
cr
ea
s
ed
,
p
r
im
ar
ily
d
u
e
to
GPU
m
em
o
r
y
an
d
p
r
o
ce
s
s
in
g
co
n
s
tr
ain
ts
,
in
d
icatin
g
t
h
e
n
ee
d
f
o
r
m
o
d
el
o
p
tim
izatio
n
o
r
h
a
r
d
war
e
au
g
m
en
tatio
n
in
lar
g
e
-
s
ca
le
d
ep
lo
y
m
e
n
ts
.
Fu
tu
r
e
wo
r
k
s
h
o
u
ld
en
h
an
ce
in
ter
p
r
etab
ilit
y
an
d
f
air
n
ess
in
L
L
aM
A
3
-
b
ased
s
cr
ee
n
in
g
th
r
o
u
g
h
ex
p
lain
a
b
le
AI
an
d
b
ias
m
itig
atio
n
tech
n
iq
u
es.
Als
o
im
p
r
o
v
i
n
g
d
ataset
d
iv
er
s
ity
an
d
d
esig
n
in
g
ef
f
ec
tiv
e
p
r
o
m
p
ts
ca
n
b
o
o
s
t
a
cc
u
r
ac
y
an
d
r
ed
u
ce
u
n
f
air
o
u
tco
m
es.
Do
m
ai
n
s
p
ec
if
ic
f
in
e
tu
n
in
g
an
d
h
u
m
a
n
in
th
e
lo
o
p
f
ee
d
b
ac
k
will
s
u
p
p
o
r
t
co
n
tin
u
o
u
s
m
o
d
el
r
ef
i
n
em
en
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
953
-
9
6
1
960
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
e
au
th
o
r
s
tates th
at
n
o
f
u
n
d
i
n
g
was in
v
o
lv
e
d
in
th
is
p
r
o
jec
t.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT
)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Asmita
Desh
m
u
k
h
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
An
jali Rau
t
✓
✓
✓
✓
✓
✓
Ved
an
t D
esh
m
u
k
h
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
Sp
ec
if
ic
d
ata
r
eq
u
ir
e
d
ca
n
b
e
m
ad
e
av
aila
b
le
u
p
o
n
r
ea
s
o
n
a
b
le
r
eq
u
est
to
th
e
c
o
r
r
esp
o
n
d
i
n
g
a
u
th
o
r
,
[
AD]
.
All
m
ater
ials
b
elo
n
g
to
th
e
au
t
h
o
r
s
a
n
d
ca
n
n
o
t
b
e
s
o
ld
to
a
n
y
o
n
e.
C
o
d
e
will
b
e
m
ad
e
av
ailab
le
u
p
o
n
r
ea
s
o
n
ab
le
r
eq
u
est to
th
e
co
r
r
e
s
p
o
n
d
in
g
au
th
o
r
.
RE
F
E
R
E
NC
E
S
[
1
]
S
.
K
.
K
o
p
p
a
r
a
p
u
,
“
A
u
t
o
m
a
t
i
c
e
x
t
r
a
c
t
i
o
n
o
f
u
s
a
b
l
e
i
n
f
o
r
ma
t
i
o
n
f
r
o
m
u
n
st
r
u
c
t
u
r
e
d
r
e
s
u
m
e
s
t
o
a
i
d
se
a
r
c
h
,
”
i
n
2
0
1
0
I
E
EE
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
Pr
o
g
ress
i
n
I
n
f
o
rm
a
t
i
c
s
a
n
d
C
o
m
p
u
t
i
n
g
,
D
e
c
.
2
0
1
0
,
p
p
.
9
9
–
1
0
3
,
d
o
i
:
1
0
.
1
1
0
9
/
P
I
C
.
2
0
1
0
.
5
6
8
7
4
2
8
.
[
2
]
A
.
B
.
C
h
a
p
ma
n
e
t
a
l
.
,
“
U
s
i
n
g
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ss
i
n
g
t
o
st
u
d
y
h
o
m
e
l
e
s
sn
e
ss
l
o
n
g
i
t
u
d
i
n
a
l
l
y
w
i
t
h
e
l
e
c
t
r
o
n
i
c
h
e
a
l
t
h
r
e
c
o
r
d
d
a
t
a
su
b
j
e
c
t
t
o
i
r
r
e
g
u
l
a
r
o
b
s
e
r
v
a
t
i
o
n
s,
”
M
a
r
.
1
8
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
1
/
2
0
2
3
.
0
3
.
1
7
.
2
3
2
8
7
4
1
4
.
[
3
]
D
.
J.
F
e
l
l
e
r
,
J.
Z
u
c
k
e
r
,
M
.
T.
Y
i
n
,
P
.
G
o
r
d
o
n
,
a
n
d
N
.
El
h
a
d
a
d
,
“
U
s
i
n
g
c
l
i
n
i
c
a
l
n
o
t
e
s
a
n
d
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ss
i
n
g
f
o
r
a
u
t
o
m
a
t
e
d
H
I
V
r
i
sk
a
ss
e
ssm
e
n
t
,
”
J
AI
D
S
J
o
u
r
n
a
l
o
f
Ac
q
u
i
re
d
I
m
m
u
n
e
D
e
f
i
c
i
e
n
c
y
S
y
n
d
ro
m
e
s
,
v
o
l
.
7
7
,
n
o
.
2
,
p
p
.
1
6
0
–
1
6
6
,
F
e
b
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
9
7
/
Q
A
I
.
0
0
0
0
0
0
0
0
0
0
0
0
1
5
8
0
.
[
4
]
J.
N
a
y
o
r
,
L
.
F
.
B
o
r
g
e
s,
S
.
G
o
r
y
a
c
h
e
v
,
V
.
S
.
G
a
i
n
e
r
,
a
n
d
J.
R
.
S
a
l
t
z
ma
n
,
“
N
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
a
c
c
u
r
a
t
e
l
y
c
a
l
c
u
l
a
t
e
s
a
d
e
n
o
ma
a
n
d
s
e
ssi
l
e
serr
a
t
e
d
p
o
l
y
p
d
e
t
e
c
t
i
o
n
r
a
t
e
s,
”
D
i
g
e
s
t
i
v
e
D
i
s
e
a
ses
a
n
d
S
c
i
e
n
c
e
s
,
v
o
l
.
6
3
,
n
o
.
7
,
p
p
.
1
7
9
4
–
1
8
0
0
,
Ju
l
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
6
2
0
-
018
-
5
0
7
8
-
4.
[
5
]
H
.
M
.
Tr
i
v
e
d
i
e
t
a
l
.
,
“
L
a
r
g
e
sca
l
e
s
e
mi
-
a
u
t
o
m
a
t
e
d
l
a
b
e
l
i
n
g
o
f
r
o
u
t
i
n
e
f
r
e
e
-
t
e
x
t
c
l
i
n
i
c
a
l
r
e
c
o
r
d
s
f
o
r
d
e
e
p
l
e
a
r
n
i
n
g
,
”
J
o
u
r
n
a
l
o
f
D
i
g
i
t
a
l
I
m
a
g
i
n
g
,
v
o
l
.
3
2
,
n
o
.
1
,
p
p
.
3
0
–
3
7
,
F
e
b
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
2
7
8
-
0
1
8
-
0
1
0
5
-
8.
[
6
]
B
.
G
u
n
a
se
e
l
a
n
,
S
.
M
a
n
d
a
l
,
a
n
d
V
.
R
a
j
a
g
o
p
a
l
a
n
,
“
A
u
t
o
m
a
t
i
c
e
x
t
r
a
c
t
i
o
n
o
f
seg
me
n
t
s
f
r
o
m
r
e
su
m
e
s
u
si
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
i
n
2
0
2
0
I
E
EE
1
7
t
h
I
n
d
i
a
C
o
u
n
c
i
l
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
(
I
N
D
I
C
O
N
)
,
D
e
c
.
2
0
2
0
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
N
D
I
C
O
N
4
9
8
7
3
.
2
0
2
0
.
9
3
4
2
5
9
6
.
[
7
]
K
.
W
a
i
l
t
h
a
r
e
,
A
.
Ta
m
h
a
n
e
,
V
.
M
u
l
i
k
,
a
n
d
K
.
S
u
r
y
a
w
a
n
s
h
,
“
A
c
o
si
n
e
si
mi
l
a
r
i
t
y
-
b
a
se
d
r
e
s
u
me
scre
e
n
i
n
g
sy
st
e
m
f
o
r
j
o
b
r
e
c
r
u
i
t
m
e
n
t
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
Re
se
a
r
c
h
J
o
u
rn
a
l
o
f
M
o
d
e
rn
i
z
a
t
i
o
n
i
n
En
g
i
n
e
e
ri
n
g
T
e
c
h
n
o
l
o
g
y
a
n
d
S
c
i
e
n
c
e
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
5
6
7
2
6
/
I
R
JM
E
TS3
5
9
4
5
.
[
8
]
P
.
K
.
R
o
y
,
S
.
S
.
C
h
o
w
d
h
a
r
y
,
a
n
d
R
.
B
h
a
t
i
a
,
“
A
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
f
o
r
a
u
t
o
ma
t
i
o
n
o
f
r
e
su
m
e
r
e
c
o
mm
e
n
d
a
t
i
o
n
sy
st
e
m,”
Pro
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
1
6
7
,
p
p
.
2
3
1
8
–
2
3
2
7
,
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s.
2
0
2
0
.
0
3
.
2
8
4
.
[
9
]
A
.
H
.
A
l
d
e
r
h
a
m
a
n
d
E.
S
.
J
a
h
a
,
“
I
mp
r
o
v
e
d
c
a
n
d
i
d
a
t
e
-
c
a
r
e
e
r
ma
t
c
h
i
n
g
u
si
n
g
c
o
m
p
a
r
a
t
i
v
e
s
e
ma
n
t
i
c
r
e
s
u
me
a
n
a
l
y
si
s,”
A
d
v
a
n
c
e
s
i
n
S
c
i
e
n
c
e
,
T
e
c
h
n
o
l
o
g
y
a
n
d
E
n
g
i
n
e
e
r
i
n
g
S
y
s
t
e
m
s
J
o
u
r
n
a
l
,
v
o
l
.
9
,
n
o
.
1
,
p
p
.
1
5
–
2
2
,
Ja
n
.
2
0
2
4
,
d
o
i
:
1
0
.
2
5
0
4
6
/
a
j
0
9
0
1
0
3
.
[
1
0
]
N
.
S
h
a
r
ma
,
R
.
B
h
u
t
i
a
,
V
.
S
a
r
d
a
r
,
A
.
P
.
G
e
o
r
g
e
,
a
n
d
F
.
A
h
me
d
,
“
N
o
v
e
l
h
i
r
i
n
g
p
r
o
c
e
ss
u
s
i
n
g
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
n
d
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
s
si
n
g
,
”
i
n
2
0
2
1
I
EE
E
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
El
e
c
t
r
o
n
i
c
s
,
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
(
C
O
N
E
C
C
T
)
,
J
u
l
.
2
0
2
1
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
C
O
N
E
C
C
T5
2
8
7
7
.
2
0
2
1
.
9
6
2
2
6
9
2
.
[
1
1
]
A
.
D
o
ss
a
t
a
y
e
v
,
A
.
M
a
n
a
p
o
v
a
,
a
n
d
B
.
O
mar
o
v
,
“
D
e
e
p
r
e
s
i
d
u
a
l
c
o
n
v
o
l
u
t
i
o
n
a
l
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
n
e
t
w
o
r
k
f
o
r
o
p
t
i
o
n
p
r
i
c
e
p
r
e
d
i
c
t
i
o
n
p
r
o
b
l
e
m,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
4
,
n
o
.
9
,
2
0
2
3
,
d
o
i
:
1
0
.
1
4
5
6
9
/
I
JA
C
S
A
.
2
0
2
3
.
0
1
4
0
9
4
1
.
[
1
2
]
S
.
B
h
a
r
a
d
w
a
j
,
R
.
V
a
r
u
n
,
P
.
S
.
A
d
i
t
y
a
,
M
.
N
i
k
h
i
l
,
a
n
d
G
.
C
.
B
a
b
u
,
“
R
e
s
u
me
s
c
r
e
e
n
i
n
g
u
si
n
g
N
LP
a
n
d
LSTM
,
”
i
n
2
0
2
2
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
v
e
n
t
i
v
e
C
o
m
p
u
t
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s
(
I
C
I
C
T
)
,
J
u
l
.
2
0
2
2
,
p
p
.
2
3
8
–
2
4
1
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
C
T5
4
3
4
4
.
2
0
2
2
.
9
8
5
0
8
8
9
.
[
1
3
]
S
.
L.
O
h
,
E.
Y
.
K
.
N
g
,
R
.
S
.
Ta
n
,
a
n
d
U
.
R
.
A
c
h
a
r
y
a
,
“
A
u
t
o
m
a
t
e
d
d
i
a
g
n
o
si
s
o
f
a
r
r
h
y
t
h
mi
a
u
s
i
n
g
c
o
m
b
i
n
a
t
i
o
n
o
f
C
N
N
a
n
d
LST
M
t
e
c
h
n
i
q
u
e
s
w
i
t
h
v
a
r
i
a
b
l
e
l
e
n
g
t
h
h
e
a
r
t
b
e
a
t
s,
”
C
o
m
p
u
t
e
rs
i
n
Bi
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
1
0
2
,
p
p
.
2
7
8
–
2
8
7
,
N
o
v
.
2
0
1
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
b
i
o
me
d
.
2
0
1
8
.
0
6
.
0
0
2
.
[
1
4
]
A
.
S
h
a
r
ma,
N
.
G
a
r
g
,
S
.
P
a
t
i
d
a
r
,
R
.
S
a
n
T
a
n
,
a
n
d
U
.
R
.
A
c
h
a
r
y
a
,
“
A
u
t
o
ma
t
e
d
p
r
e
-
scr
e
e
n
i
n
g
o
f
a
r
r
h
y
t
h
m
i
a
u
si
n
g
h
y
b
r
i
d
c
o
m
b
i
n
a
t
i
o
n
o
f
F
o
u
r
i
e
r
–
B
e
ss
e
l
e
x
p
a
n
si
o
n
a
n
d
LS
TM
,
”
C
o
m
p
u
t
e
rs
i
n
Bi
o
l
o
g
y
a
n
d
Me
d
i
c
i
n
e
,
v
o
l
.
1
2
0
,
M
a
y
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
m
p
b
i
o
me
d
.
2
0
2
0
.
1
0
3
7
5
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ca
la
b
le
r
esu
me
s
creen
in
g
u
s
i
n
g
la
r
g
e
la
n
g
u
a
g
e
mo
d
el
Meta
A
I
ve
r
s
io
n
3
(
A
s
mita
Desh
mu
kh
)
961
[
1
5
]
C
.
A
t
h
u
k
o
r
a
l
a
,
H
.
K
u
mar
a
si
n
g
h
e
,
K
.
D
a
b
a
r
e
,
P
.
U
j
i
t
h
a
n
g
a
n
a
,
S
.
Th
e
l
i
j
j
a
g
o
d
a
,
a
n
d
P
.
L
i
y
a
n
a
g
e
,
“
B
u
s
i
n
e
ss
i
n
t
e
l
l
i
g
e
n
c
e
a
ssi
s
t
a
n
t
f
o
r
h
u
ma
n
r
e
s
o
u
r
c
e
ma
n
a
g
e
m
e
n
t
f
o
r
I
T
c
o
mp
a
n
i
e
s,
”
i
n
2
0
2
0
2
0
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
A
d
v
a
n
c
e
s
i
n
I
C
T
f
o
r
Em
e
r
g
i
n
g
Re
g
i
o
n
s
(
I
C
T
e
r)
,
N
o
v
.
2
0
2
0
,
p
p
.
2
2
0
–
2
2
5
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
Te
r
5
1
0
9
7
.
2
0
2
0
.
9
3
2
5
4
7
1
.
[
1
6
]
G
.
L,
R
.
M
L,
G
.
H
B
,
K
.
M
a
t
h
a
d
a
,
a
n
d
M
.
B
,
“
I
n
t
e
l
l
i
g
e
n
t
r
e
s
u
me
scru
t
i
n
y
u
si
n
g
n
a
m
e
d
e
n
t
i
t
y
r
e
c
o
g
n
i
t
i
o
n
w
i
t
h
B
ER
T,
”
i
n
2
0
2
3
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
D
a
t
a
S
c
i
e
n
c
e
a
n
d
N
e
t
w
o
rk
S
e
c
u
r
i
t
y
(
I
C
D
S
N
S
)
,
Ju
l
.
2
0
2
3
,
p
p
.
0
1
–
08
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
D
S
N
S
5
8
4
6
9
.
2
0
2
3
.
1
0
2
4
5
3
0
4
.
[
1
7
]
E.
A
b
d
o
l
l
a
h
n
e
j
a
d
,
M
.
K
a
l
m
a
n
,
a
n
d
B
.
H
.
F
a
r
,
“
A
d
e
e
p
l
e
a
r
n
i
n
g
B
ER
T
-
b
a
s
e
d
a
p
p
r
o
a
c
h
t
o
p
e
r
so
n
-
j
o
b
f
i
t
i
n
t
a
l
e
n
t
r
e
c
r
u
i
t
m
e
n
t
,
”
i
n
2
0
2
1
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
C
o
m
p
u
t
a
t
i
o
n
a
l
S
c
i
e
n
c
e
a
n
d
C
o
m
p
u
t
a
t
i
o
n
a
l
I
n
t
e
l
l
i
g
e
n
c
e
(
C
S
C
I
)
,
D
e
c
.
2
0
2
1
,
p
p
.
9
8
–
1
0
4
,
d
o
i
:
1
0
.
1
1
0
9
/
C
S
C
I
5
4
9
2
6
.
2
0
2
1
.
0
0
0
9
1
.
[
1
8
]
M
.
B
.
P
a
d
m
a
j
a
,
C
.
S
.
K
u
m
a
r
,
V
.
P
r
e
e
t
h
i
,
B
.
S
.
S
o
m
e
sw
a
r
,
a
n
d
A
.
G
a
n
t
a
y
a
t
,
“
A
w
e
b
a
p
p
l
i
c
a
t
i
o
n
-
b
a
s
e
d
mo
c
k
i
n
t
e
r
v
i
e
w
u
s
i
n
g
N
LP
t
e
c
h
n
i
q
u
e
s,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Ad
v
a
n
c
e
d
R
e
se
a
rc
h
i
n
E
d
u
c
a
t
i
o
n
a
n
d
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
1
,
n
o
.
6
,
p
p
.
2
6
6
9
–
2
6
7
7
,
2
0
2
4
,
d
o
i
:
1
0
.
1
5
6
8
0
/
I
JA
R
ETY
.
2
0
2
4
.
1
1
0
6
0
1
6
.
[
1
9
]
S
.
A
.
P
i
a
s,
M
.
H
o
ss
a
i
n
,
H
.
R
a
h
ma
n
,
a
n
d
M
d
.
M
.
H
o
ss
a
i
n
,
“
E
n
h
a
n
c
i
n
g
j
o
b
m
a
t
c
h
i
n
g
t
h
r
o
u
g
h
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ss
i
n
g
:
a
B
ER
T
-
b
a
se
d
a
p
p
r
o
a
c
h
,
”
i
n
2
0
2
4
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
I
n
n
o
v
a
t
i
o
n
s
i
n
S
c
i
e
n
c
e
,
E
n
g
i
n
e
e
r
i
n
g
a
n
d
T
e
c
h
n
o
l
o
g
y
(
I
C
I
S
ET)
,
O
c
t
.
2
0
2
4
,
p
p
.
1
–
6
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
I
S
ET6
2
1
2
3
.
2
0
2
4
.
1
0
9
3
9
8
6
0
.
[
2
0
]
V
.
Jame
s,
A
.
K
u
l
k
a
r
n
i
,
a
n
d
R
.
A
g
a
r
w
a
l
,
“
R
e
s
u
me
s
h
o
r
t
l
i
st
i
n
g
a
n
d
r
a
n
k
i
n
g
w
i
t
h
t
r
a
n
sf
o
r
mers
,
”
i
n
I
n
t
e
l
l
i
g
e
n
t
S
y
st
e
m
s
a
n
d
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
4
7
1
,
S
.
N
a
n
d
a
n
M
o
h
a
n
t
y
,
V
.
G
a
r
c
i
a
D
i
a
z
,
a
n
d
G
.
A
.
E.
S
a
t
i
sh
K
u
mar
,
E
d
s.,
C
h
a
m
:
S
p
r
i
n
g
e
r
N
a
t
u
r
e
S
w
i
t
z
e
r
l
a
n
d
,
2
0
2
3
,
p
p
.
9
9
–
1
0
8
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
1
-
3
5
0
8
1
-
8
_
8
.
[
2
1
]
M
.
K
a
y
a
a
n
d
T.
B
o
g
e
r
s,
“
A
n
e
x
p
l
o
r
a
t
i
o
n
o
f
se
n
t
e
n
c
e
-
p
a
i
r
c
l
a
ssi
f
i
c
a
t
i
o
n
f
o
r
a
l
g
o
r
i
t
h
m
i
c
r
e
c
r
u
i
t
i
n
g
,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
1
7
t
h
AC
M
C
o
n
f
e
r
e
n
c
e
o
n
Re
c
o
m
m
e
n
d
e
r
S
y
st
e
m
s
,
S
e
p
.
2
0
2
3
,
p
p
.
1
1
7
5
–
1
1
7
9
,
d
o
i
:
1
0
.
1
1
4
5
/
3
6
0
4
9
1
5
.
3
6
1
0
6
5
7
.
[
2
2
]
M
.
L
i
,
J.
S
u
n
,
a
n
d
X
.
Ta
n
,
“
E
v
a
l
u
a
t
i
n
g
t
h
e
e
f
f
e
c
t
i
v
e
n
e
ss
o
f
l
a
r
g
e
l
a
n
g
u
a
g
e
mo
d
e
l
s
i
n
a
b
s
t
r
a
c
t
scre
e
n
i
n
g
:
a
c
o
mp
a
r
a
t
i
v
e
a
n
a
l
y
s
i
s,”
S
y
s
t
e
m
a
t
i
c
R
e
v
i
e
w
s
,
v
o
l
.
1
3
,
n
o
.
1
,
A
u
g
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s
1
3
6
4
3
-
0
2
4
-
0
2
6
0
9
-
x.
[
2
3
]
C
.
G
a
n
,
Q
.
Z
h
a
n
g
,
a
n
d
T.
M
o
r
i
,
“
A
p
p
l
i
c
a
t
i
o
n
o
f
LL
M
a
g
e
n
t
s
i
n
r
e
c
r
u
i
t
me
n
t
:
a
n
o
v
e
l
f
r
a
mew
o
r
k
f
o
r
a
u
t
o
ma
t
e
d
r
e
su
me
scre
e
n
i
n
g
,
”
J
o
u
rn
a
l
o
f
I
n
f
o
rm
a
t
i
o
n
P
ro
c
e
ss
i
n
g
,
v
o
l
.
3
2
,
p
p
.
8
8
1
–
8
9
3
,
2
0
2
4
,
d
o
i
:
1
0
.
2
1
9
7
/
i
p
s
j
j
i
p
.
3
2
.
8
8
1
.
[
2
4
]
P
.
P
a
t
i
l
,
A
.
K
h
a
r
d
e
,
S
.
D
e
o
c
h
a
k
e
,
a
n
d
V
.
K
h
a
r
a
t
,
“
A
p
p
l
i
c
a
t
i
o
n
o
f
R
A
G
(
r
e
t
r
i
e
v
a
l
-
a
u
g
me
n
t
e
d
g
e
n
e
r
a
t
i
o
n
)
i
n
A
I
-
d
r
i
v
e
n
r
e
s
u
me
a
n
a
l
y
si
s
a
n
d
j
o
b
ma
t
c
h
i
n
g
,
”
I
n
t
e
rn
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
R
e
se
a
rc
h
i
n
S
c
i
e
n
c
e
,
C
o
m
m
u
n
i
c
a
t
i
o
n
a
n
d
T
e
c
h
n
o
l
o
g
y
,
p
p
.
6
7
1
–
6
7
9
,
M
a
r
.
2
0
2
5
,
d
o
i
:
1
0
.
4
8
1
7
5
/
I
JA
R
S
C
T
-
2
5
1
8
9
.
[
2
5
]
F
.
P
.
-
W
.
Lo
e
t
a
l
.
,
“
A
I
h
i
r
i
n
g
w
i
t
h
L
LM
s
:
a
c
o
n
t
e
x
t
-
a
w
a
r
e
a
n
d
e
x
p
l
a
i
n
a
b
l
e
m
u
l
t
i
-
a
g
e
n
t
f
r
a
mew
o
r
k
f
o
r
r
e
s
u
m
e
scree
n
i
n
g
,
”
2
0
2
5
,
a
rXi
v
:
2
5
0
4
.
0
2
8
7
0
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
As
m
ita
De
shm
u
k
h
is
a
ss
istan
t
p
r
o
fe
ss
o
r
a
t
KJ
S
I
T,
M
u
m
b
a
i,
In
d
ia.
S
h
e
is
c
u
rre
n
tl
y
p
u
rsu
i
n
g
P
h
.
D.
(Co
m
p
u
ter
S
c
ien
c
e
En
g
in
e
e
rin
g
)
fro
m
S
a
n
t
G
a
d
g
e
Ba
b
a
Am
ra
v
a
ti
Un
iv
e
rsity
,
m
a
ste
r
o
f
e
n
g
in
e
e
ri
n
g
(C
o
m
p
u
ter
En
g
i
n
e
e
rin
g
)
fro
m
Un
iv
e
rsity
o
f
M
u
m
b
a
i
a
n
d
b
a
c
h
e
lo
r
o
f
e
n
g
i
n
e
e
rin
g
(C
o
m
p
u
t
e
r
S
c
ien
c
e
En
g
in
e
e
rin
g
)
fro
m
A
m
ra
v
a
ti
Un
iv
e
rsity
i
n
1
9
9
6
a
n
d
2
0
1
2
.
He
r
p
rima
ry
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
a
n
d
d
e
e
p
lea
rn
in
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
sm
it
a
d
e
sh
m
u
k
h
7
@
g
m
a
il
.
c
o
m
.
Anja
li
R
a
u
t
is
p
ro
fe
ss
o
r
a
n
d
i
n
c
h
a
rg
e
p
rin
c
i
p
a
l
a
t
HV
P
M
Co
l
leg
e
o
f
En
g
in
e
e
rin
g
a
n
d
Tec
h
n
o
l
o
g
y
,
Am
ra
v
a
ti
In
d
ia.
S
h
e
h
a
s
c
o
m
p
lete
d
h
e
r
P
h
.
D.
(C
o
m
p
u
ter
S
c
ien
c
e
En
g
i
n
e
e
rin
g
)
fro
m
S
a
n
t
G
a
d
g
e
Ba
b
a
Am
ra
v
a
ti
Un
iv
e
rsity
,
m
a
ste
r
o
f
e
n
g
i
n
e
e
rin
g
(Co
m
p
u
ter
S
c
ien
c
e
En
g
in
e
e
rin
g
)
a
n
d
b
a
c
h
e
lo
r
o
f
e
n
g
i
n
e
e
rin
g
(Co
m
p
u
ter
S
c
ien
c
e
En
g
i
n
e
e
rin
g
)
fr
o
m
Am
ra
v
a
ti
Un
iv
e
rsity
in
1
9
9
4
a
n
d
2
0
1
3
.
He
r
m
a
in
a
re
a
o
f
re
se
a
rc
h
i
n
tere
sts
is
d
a
ta
m
in
i
n
g
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
n
jali_
d
a
h
a
k
e
@re
d
iffma
il
.
c
o
m
.
Ve
d
a
n
t
De
shm
u
k
h
is
a
fin
a
l
-
y
e
a
r
stu
d
e
n
t
a
t
t
h
e
De
p
a
rtme
n
t
o
f
C
o
m
p
u
ter
En
g
i
n
e
e
rin
g
,
S
a
rd
a
r
P
a
tel
In
sti
tu
te
o
f
Tec
h
n
o
l
o
g
y
(S
P
IT)
,
M
u
m
b
a
i,
In
d
ia.
He
is
e
x
p
e
c
ted
to
re
c
e
iv
e
h
is
B.
Tec
h
.
d
e
g
re
e
i
n
C
o
m
p
u
ter
En
g
in
e
e
rin
g
i
n
2
0
2
5
.
H
is
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
m
e
d
ica
l
ima
g
in
g
u
si
n
g
c
o
m
p
u
ter
v
isi
o
n
a
n
d
m
a
c
h
in
e
lea
rn
i
n
g
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
v
e
d
a
n
t
d
e
sh
m
u
k
h
3
1
0
8
@g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.