I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
39
,
No
.
1
,
J
u
l
y
2
0
2
5
,
p
p
.
446
~
4
5
4
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
:
1
0
.
1
1
5
9
1
/ijeecs.v
39
.i
1
.
pp
446
-
4
5
4
446
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
Inv
estig
a
ting
the
recall ef
ficiency
i
n abs
tractive sum
ma
riza
tion:
a
n ex
pe
rimental
ba
sed co
mpa
ra
tive study
Su
ra
bh
i A
nu
ra
dh
a
1
,
2
,
M
a
rt
h
a
Sh
esh
ik
a
la
1
1
S
c
h
o
o
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
S
R
U
n
i
v
e
r
s
i
t
y
,
W
a
r
a
n
g
a
l
,
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
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
(
A
I
M
L)
,
K
e
sh
a
v
M
e
m
o
r
i
a
l
I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
H
y
d
e
r
a
b
a
d
,
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
Au
g
5
,
2
0
2
4
R
ev
is
ed
Mar
17
,
2
0
2
5
Acc
ep
ted
Mar
26
,
2
0
2
5
Th
is
st
u
d
y
e
x
p
l
o
re
s
te
x
t
s
u
m
m
a
riza
ti
o
n
,
a
c
rit
ica
l
c
o
m
p
o
n
e
n
t
o
f
n
a
tu
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
i
n
g
(NLP
),
sp
e
c
ifi
c
a
ll
y
targ
e
ti
n
g
sc
ien
ti
fic
d
o
c
u
m
e
n
ts.
Trad
it
io
n
a
l
e
x
trac
ti
v
e
su
m
m
a
riza
ti
o
n
,
wh
ich
re
li
e
s
o
n
t
h
e
o
ri
g
in
a
l
wo
rd
in
g
,
o
ften
re
su
l
ts
in
d
isj
o
in
te
d
se
q
u
e
n
c
e
s
o
f
se
n
ten
c
e
s
a
n
d
fa
il
s
to
c
o
n
v
e
y
k
e
y
id
e
a
s
c
o
n
c
ise
ly
.
To
a
d
d
re
ss
t
h
e
se
issu
e
s
a
n
d
e
n
s
u
re
c
o
m
p
re
h
e
n
siv
e
in
c
lu
sio
n
o
f
re
lev
a
n
t
d
e
tails,
o
u
r
re
se
a
rc
h
a
ims
to
imp
r
o
v
e
th
e
c
o
h
e
re
n
c
e
a
n
d
c
o
m
p
lete
n
e
ss
o
f
su
m
m
a
ries
.
We
e
m
p
lo
y
e
d
2
5
d
iffere
n
t
lar
g
e
lan
g
u
a
g
e
m
o
d
e
ls
(LL
M
s)
t
o
e
v
a
lu
a
te
th
e
ir
p
e
rf
o
rm
a
n
c
e
in
g
e
n
e
ra
ti
n
g
a
b
stra
c
ti
v
e
su
m
m
a
ries
o
f
sc
h
o
larly
sc
ien
ti
fic
d
o
c
u
m
e
n
ts.
A
re
c
a
ll
-
o
rien
ted
e
v
a
lu
a
ti
o
n
o
f
th
e
g
e
n
e
ra
ted
s
u
m
m
a
ries
re
v
e
a
le
d
th
a
t
LL
M
s
su
c
h
a
s
'
Clau
d
e
v
2
.
1
,
'
'
P
P
LX
7
0
B
O
n
li
n
e
,
'
a
n
d
'
M
istral
7
B
In
st
ru
c
t'
d
e
m
o
n
stra
ted
e
x
c
e
p
ti
o
n
a
l
p
e
rfo
rm
a
n
c
e
with
ROU
G
E
-
1
sc
o
re
s
o
f
0
.
9
2
,
0
.
8
8
,
a
n
d
0
.
8
5
,
re
sp
e
c
ti
v
e
l
y
,
su
p
p
o
rted
b
y
h
ig
h
p
re
c
isio
n
a
n
d
re
c
a
ll
v
a
lu
e
s
fro
m
b
id
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tatio
n
s
fro
m
tran
sfo
rm
e
rs
(
BERT
)
sc
o
re
s
(0
.
9
0
2
,
0
.
8
9
4
,
a
n
d
0
.
8
8
8
).
Th
e
s
e
fin
d
in
g
s
o
ffe
r
v
a
lu
a
b
le
in
si
g
h
ts
f
o
r
NL
P
r
e
se
a
rc
h
e
rs,
lay
in
g
t
h
e
fo
u
n
d
a
ti
o
n
fo
r
fu
tu
re
a
d
v
a
n
c
e
m
e
n
ts
in
LL
M
s
fo
r
su
m
m
a
riza
ti
o
n
.
T
h
e
stu
d
y
h
i
g
h
l
ig
h
ts
p
o
ten
ti
a
l
imp
ro
v
e
m
e
n
ts
i
n
tex
t
s
u
m
m
a
riza
ti
o
n
tec
h
n
iq
u
e
s,
b
e
n
e
fi
ti
n
g
v
a
r
io
u
s
NLP
a
p
p
li
c
a
ti
o
n
s.
K
ey
w
o
r
d
s
:
Ab
s
tr
ac
tiv
e
s
u
m
m
ar
y
C
lau
d
e
C
o
h
er
en
ce
Falco
n
-
7B
-
in
s
tr
u
ct
L
ar
g
e
lan
g
u
ag
e
m
o
d
els
Mistra
l
-
7B
My
th
o
Ma
x
-
13B
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
:
Su
r
ab
h
i A
n
u
r
ad
h
a
Sch
o
o
l o
f
C
o
m
p
u
ter
Scien
ce
a
n
d
Ar
tific
i
al
I
n
tellig
en
ce
,
SR
Un
iv
er
s
ity
W
ar
an
g
al
,
I
n
d
ia
E
m
ail:
an
u
r
ad
h
a@
k
m
it.in
1.
I
NT
RO
D
UCT
I
O
N
T
ex
t
s
u
m
m
ar
izatio
n
is
ess
en
tial
in
to
d
ay
'
s
d
ata
-
d
r
iv
en
wo
r
ld
,
esp
ec
ially
in
f
ield
s
lik
e
jo
u
r
n
alis
m
,
law,
an
d
ac
ad
em
ia.
I
n
jo
u
r
n
alis
m
,
ef
f
icien
t
s
u
m
m
ar
izatio
n
en
ab
les
q
u
ick
d
ig
esti
o
n
o
f
v
ast
am
o
u
n
ts
o
f
in
f
o
r
m
atio
n
,
f
ac
ilit
atin
g
tim
ely
n
ews u
p
d
ates.
L
eg
al
p
r
o
f
ess
io
n
als r
ely
o
n
s
u
m
m
ar
izatio
n
to
ex
tr
ac
t k
ey
p
o
in
ts
f
r
o
m
le
n
g
th
y
d
o
cu
m
en
ts
,
ai
d
in
g
in
ca
s
e
p
r
e
p
ar
atio
n
an
d
r
esear
ch
.
Similar
ly
,
in
ac
ad
e
m
ia,
s
u
m
m
ar
izatio
n
h
elp
s
r
esear
ch
er
s
s
if
t th
r
o
u
g
h
ex
ten
s
iv
e
liter
atu
r
e,
id
e
n
tify
in
g
r
elev
an
t stu
d
ies ef
f
icien
tl
y
.
T
ex
t
s
u
m
m
ar
izatio
n
f
o
c
u
s
es
o
n
co
n
d
en
s
in
g
a
g
i
v
en
te
x
t
wh
i
le
r
etain
in
g
its
ess
en
tial
in
f
o
r
m
atio
n
an
d
m
ain
id
ea
s
.
T
h
e
g
o
al
is
to
p
r
o
d
u
ce
a
co
n
cise
an
d
co
h
er
en
t
s
u
m
m
ar
y
th
at
ac
c
u
r
ately
r
ef
lec
ts
th
e
co
r
e
m
ea
n
in
g
o
f
th
e
o
r
ig
in
al
co
n
ten
t.
B
r
o
ad
ly
,
tex
t
s
u
m
m
ar
izatio
n
ca
n
b
e
class
if
ied
in
to
two
m
ain
ap
p
r
o
ac
h
es:
ab
s
tr
ac
tiv
e
an
d
ex
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
.
E
x
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
in
v
o
lv
es
s
elec
tin
g
k
ey
s
en
ten
ce
s
o
r
p
h
r
ases
d
ir
ec
tly
f
r
o
m
th
e
o
r
ig
in
al
tex
t
to
f
o
r
m
a
s
u
m
m
ar
y
,
as
n
o
te
d
b
y
Gh
a
d
im
i
an
d
B
eig
y
[
1
]
.
T
h
is
m
et
h
o
d
ex
tr
ac
ts
co
n
ten
t
v
er
b
atim
,
em
p
h
asizin
g
im
p
o
r
t
an
ce
o
r
r
elev
an
ce
cr
iter
ia.
Al
g
o
r
ith
m
s
ass
ess
s
en
ten
ce
s
b
as
ed
o
n
f
ea
tu
r
es
lik
e
len
g
th
,
wo
r
d
f
r
eq
u
en
cy
,
an
d
k
ey
k
ey
wo
r
d
s
,
f
ac
ilit
atin
g
t
h
e
cr
ea
tio
n
o
f
a
s
u
cc
in
ct
r
e
p
r
esen
tatio
n
o
f
th
e
o
r
ig
in
al
m
ater
ial.
C
o
n
v
er
s
ely
,
ab
s
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
is
a
m
o
r
e
ad
v
a
n
ce
d
ap
p
r
o
ac
h
th
at
d
o
es
n
o
t
m
er
ely
ex
tr
ac
t
s
en
ten
ce
s
f
r
o
m
th
e
o
r
i
g
in
al
tex
t
b
u
t
in
s
tead
g
en
er
ate
s
en
tire
ly
n
ew
o
n
es.
T
h
is
m
eth
o
d
d
em
a
n
d
s
a
d
ee
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
I
n
ve
s
tig
a
tin
g
th
e
r
ec
a
ll e
fficie
n
cy
in
a
b
s
tr
a
ctive
s
u
mma
r
iz
a
t
io
n
:
a
n
ex
p
erimen
ta
l
…
(
S
u
r
a
b
h
i A
n
u
r
a
d
h
a
)
447
u
n
d
er
s
tan
d
i
n
g
o
f
th
e
co
n
ten
t
t
o
cr
ea
te
a
s
u
m
m
ar
y
th
at
is
b
o
t
h
co
h
e
r
en
t
a
n
d
n
atu
r
al
-
s
o
u
n
d
i
n
g
.
T
o
ac
co
m
p
lis
h
th
is
,
s
o
p
h
is
ticated
m
o
d
els,
p
ar
ticu
lar
ly
lar
g
e
lan
g
u
ag
e
m
o
d
e
ls
(
L
L
Ms)
,
ar
e
u
tili
ze
d
.
As
s
tated
b
y
B
asy
al
an
d
San
g
h
v
i
[
2
]
th
ese
L
L
Ms
h
av
e
th
e
ca
p
ac
ity
to
r
ewr
ite
an
d
r
ep
h
r
ase
co
n
ten
t,
ca
p
tu
r
in
g
th
e
ess
en
ce
o
f
th
e
o
r
ig
in
al
tex
t
in
a
m
o
r
e
co
n
ci
s
e
an
d
n
u
an
ce
d
m
an
n
er
.
T
h
e
u
s
e
o
f
L
L
Ms
co
n
tr
ib
u
tes
to
th
e
d
ev
elo
p
m
e
n
t
o
f
s
u
m
m
ar
ies
th
at
n
o
t
o
n
ly
c
o
n
v
ey
th
e
co
r
e
in
f
o
r
m
atio
n
b
u
t
al
s
o
ex
h
ib
it
a
h
ig
h
e
r
lev
el
o
f
lin
g
u
is
tic
f
lu
en
cy
an
d
co
n
tex
t a
war
en
ess
.
I
n
r
ec
en
t
tim
es,
tex
t
s
u
m
m
ar
izatio
n
h
as
u
n
d
er
g
o
n
e
tr
an
s
f
o
r
m
ativ
e
ch
an
g
es,
p
r
im
ar
ily
p
r
o
p
elled
b
y
th
e
r
is
e
o
f
L
L
Ms.
Pro
m
in
en
t m
o
d
els
lik
e
Op
en
AI
'
s
C
h
atG
P
T
,
MPT
-
7b
-
in
s
tr
u
ct,
f
la
n
-
t5
-
x
l,
f
alco
n
-
7b
-
in
s
tr
u
ct
,
Mistra
l
-
7b
-
in
s
tr
u
ct,
My
th
o
m
is
t,
L
lam
a,
C
lau
d
a,
PaL
M,
Her
m
es,
T
o
p
p
y
M
7
B
,
PP
L
X
On
lin
e
L
L
Ms
h
av
e
em
er
g
ed
as
p
io
n
ee
r
s
,
d
is
p
lay
in
g
im
p
r
ess
iv
e
ab
ilit
ies
in
u
n
d
er
s
tan
d
in
g
an
d
g
e
n
er
atin
g
tex
t
with
h
u
m
an
-
lik
e
f
lu
en
cy
.
T
h
is
p
ar
ad
ig
m
s
h
if
t
h
as
r
aised
th
e
b
ar
an
d
p
a
v
ed
th
e
way
f
o
r
u
n
p
r
ec
ed
e
n
te
d
p
r
o
g
r
ess
in
tex
t
s
u
m
m
ar
izatio
n
.
T
h
e
ar
ticle
e
x
p
lo
r
es
th
e
p
r
o
f
o
u
n
d
im
p
ac
t
o
f
th
ese
L
L
Ms,
h
ig
h
lig
h
tin
g
th
e
ir
p
o
ten
t
g
en
e
r
ativ
e
ca
p
ab
ilit
ies
an
d
ad
ap
tab
ilit
y
ac
r
o
s
s
d
iv
er
s
e
task
s
th
r
o
u
g
h
f
in
e
-
tu
n
in
g
.
T
h
e
e
x
am
in
atio
n
o
f
th
ese
m
o
d
els
u
n
v
eils
n
ew
p
o
s
s
ib
ilit
ies
f
o
r
e
n
h
an
cin
g
tex
t
s
u
m
m
ar
izatio
n
m
eth
o
d
o
l
o
g
ies,
m
ar
k
in
g
a
p
iv
o
tal
m
o
m
en
t
in
th
e
ev
o
lu
tio
n
o
f
NL
P tec
h
n
o
lo
g
ie
s
.
Hav
in
g
its
r
o
o
ts
in
E
n
g
lis
h
la
n
g
u
ag
e
p
r
e
-
tr
ai
n
in
g
,
th
e
b
id
ir
e
ctio
n
al
an
d
au
t
o
-
r
eg
r
ess
iv
e
tr
an
s
f
o
r
m
er
s
(
B
AR
T
)
m
o
d
el
h
as
u
n
d
er
g
o
n
e
a
tr
an
s
f
o
r
m
ativ
e
f
in
e
-
t
u
n
in
g
p
r
o
ce
s
s
o
n
th
e
C
NN
D
aily
Ma
il
d
ataset.
Op
er
atin
g
as
a
tr
a
n
s
f
o
r
m
er
-
b
a
s
ed
en
co
d
er
-
d
ec
o
d
er
m
o
d
el
w
ith
a
b
id
ir
ec
tio
n
al
en
co
d
er
a
k
in
to
b
id
ir
ec
tio
n
al
en
co
d
er
r
ep
r
esen
tatio
n
s
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
an
d
an
au
to
r
eg
r
ess
iv
e
d
ec
o
d
er
r
esem
b
lin
g
GPT,
B
AR
T
em
er
g
es
as
a
v
er
s
atile
p
o
wer
h
o
u
s
e.
I
ts
ex
ce
llen
ce
ex
ten
d
s
to
tex
t
g
en
er
atio
n
task
s
lik
e
s
u
m
m
ar
izatio
n
an
d
tr
an
s
latio
n
,
wh
ile
also
p
r
o
v
i
n
g
ad
ep
t
in
co
m
p
r
eh
e
n
s
io
n
task
s
s
u
ch
as
tex
t
c
lass
if
icatio
n
an
d
q
u
esti
o
n
an
s
wer
in
g
.
T
h
is
s
p
ec
if
ic
iter
a
tio
n
,
‘
f
ac
eb
o
o
k
/b
a
r
t
-
lar
g
e
-
c
n
n
’
,
f
in
e
-
tu
n
e
d
o
n
th
e
ex
ten
s
iv
e
C
NN
Daily
Ma
il
d
ataset,
f
u
r
th
er
a
m
p
lifie
s
B
AR
T
'
s
p
r
o
f
icien
cy
,
lev
e
r
ag
in
g
a
v
ast
co
llectio
n
o
f
tex
t
-
s
u
m
m
a
r
y
p
air
s
to
r
ef
in
e
its
lan
g
u
ag
e
u
n
d
e
r
s
tan
d
in
g
a
n
d
g
en
er
atio
n
ca
p
a
b
ilit
ies [
3
]
.
T
h
e
'
tex
t
-
d
av
in
ci
-
0
0
3
(
L
eg
ac
y
)
'
m
o
d
el
r
ep
r
esen
ts
a
s
ig
n
if
ican
t
ad
v
an
ce
m
e
n
t
in
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P
)
,
d
em
o
n
s
tr
atin
g
ex
ce
p
tio
n
al
ac
cu
r
ac
y
an
d
p
r
o
f
icien
cy
ac
r
o
s
s
a
wid
e
r
an
g
e
o
f
lan
g
u
ag
e
task
s
.
I
t
s
u
r
p
ass
es
its
p
r
ed
ec
ess
o
r
s
,
s
u
ch
as
th
e
C
u
r
ie,
B
ab
b
ag
e,
an
d
Ad
a
m
o
d
els,
b
y
co
n
s
is
ten
tly
p
r
o
d
u
ci
n
g
h
ig
h
er
-
q
u
ality
an
d
m
o
r
e
ex
te
n
d
ed
tex
t
o
u
tp
u
ts
wh
ile
ad
h
er
in
g
clo
s
ely
to
g
iv
en
in
s
tr
u
ct
io
n
s
.
W
ith
a
to
k
en
ca
p
ac
ity
o
f
4
,
0
9
7
,
th
is
leg
ac
y
m
o
d
el
ef
f
icien
tly
h
an
d
les
ex
t
en
s
iv
e
tex
t
g
e
n
er
atio
n
[
4
]
.
No
tab
ly
,
‘
GPT
-
4
’
h
as
b
ee
n
r
ec
o
g
n
ize
d
f
o
r
its
im
p
r
o
v
em
en
ts
o
v
er
‘
GPT
-
3
.
5
’
Op
en
AI
em
p
h
asizes
th
at
‘
GPT
-
4
’
is
m
o
r
e
r
eliab
le,
cr
ea
tiv
e,
an
d
ca
p
a
b
le
o
f
h
a
n
d
l
in
g
m
o
r
e
n
u
an
ce
d
i
n
s
tr
u
ctio
n
s
.
No
tab
ly
,
‘
GPT
-
4
’
in
tr
o
d
u
ce
s
two
v
er
s
io
n
s
with
s
ig
n
if
ican
tly
ex
p
an
d
e
d
co
n
tex
t
win
d
o
ws,
allo
win
g
f
o
r
th
e
p
r
o
ce
s
s
in
g
o
f
8
,
1
9
2
an
d
3
2
,
7
6
8
to
k
en
s
.
T
h
is
m
ar
k
s
a
s
u
b
s
tan
tial
im
p
r
o
v
em
en
t
co
m
p
ar
ed
to
t
h
e
lim
itatio
n
s
o
f
‘
GPT
-
3
.
5
’
a
n
d
‘
GPT
-
3
’
,
wh
ic
h
wer
e
c
o
n
f
i
n
ed
t
o
4
,
0
9
6
an
d
2
,
0
4
9
to
k
en
s
,
r
esp
e
ctiv
ely
[
5
]
.
As
s
tated
in
r
e
f
er
en
ce
[
6
]
,
t
h
e
‘
MPT
-
7B
-
I
n
s
tr
u
ct’
m
o
d
el
is
d
esig
n
ed
s
p
ec
if
ically
f
o
r
s
h
o
r
t
-
f
o
r
m
in
s
tr
u
ctio
n
-
f
o
llo
win
g
task
s
,
m
ak
in
g
it
an
ex
ce
lle
n
t
ch
o
ic
e
f
o
r
v
ar
i
o
u
s
in
s
tr
u
ctio
n
-
b
ased
ap
p
licatio
n
s
.
T
h
is
7
-
b
illi
o
n
-
p
ar
am
eter
LLM
wa
s
tr
ain
ed
o
n
1
tr
illi
o
n
to
k
e
n
s
o
v
er
9
.
5
d
ay
s
u
s
in
g
4
4
0
A1
0
0
-
4
0
G
g
r
a
p
h
ics
p
r
o
ce
s
s
in
g
u
n
its
(
GPUs
)
.
I
t is d
ev
elo
p
e
d
b
y
f
in
e
-
tu
n
in
g
th
e
b
ase
m
o
d
el,
‘
MPT
-
7
B
,
’
with
th
e
an
th
r
o
p
ic
h
elp
f
u
l
an
d
h
ar
m
less
(
HH
-
R
L
HF)
d
ataset
an
d
Data
b
r
ick
s
Do
lly
-
1
5
k
d
ataset.
T
h
is
tailo
r
ed
a
p
p
r
o
ac
h
r
esu
lts
in
a
m
o
d
el
th
at
ex
ce
ls
at
ac
cu
r
ately
co
m
p
r
eh
en
d
in
g
an
d
f
o
ll
o
win
g
in
s
tr
u
ctio
n
s
with
p
r
ec
is
io
n
.
Featu
r
in
g
a
d
ec
o
d
er
-
o
n
ly
a
r
ch
itectu
r
e,
t
h
e
m
o
d
el
is
o
p
tim
ized
f
o
r
s
ce
n
ar
i
o
s
wh
er
e
u
s
er
s
ask
q
u
esti
o
n
s
a
n
d
ex
p
ec
t c
o
n
cise,
d
ir
ec
t r
esp
o
n
s
es r
ath
er
t
h
an
an
ex
ten
d
ed
co
n
tin
u
atio
n
o
f
th
eir
in
p
u
t.
‘
Falco
n
-
7B
-
I
n
s
tr
u
ct’
is
a
h
ig
h
ly
ca
p
ab
le
7
-
b
illi
o
n
-
p
a
r
am
eter
ca
u
s
al
d
ec
o
d
er
-
o
n
l
y
m
o
d
el,
m
eticu
lo
u
s
ly
d
e
v
elo
p
e
d
b
y
th
e
T
ec
h
n
o
lo
g
y
I
n
n
o
v
atio
n
I
n
s
t
itu
te
(
T
I
I
)
as
a
n
e
x
ten
s
io
n
o
f
‘
Falco
n
-
7
B
’
.
Fin
e
-
tu
n
ed
o
n
a
d
iv
er
s
e
d
ataset
f
r
o
m
ch
at
an
d
in
s
tr
u
ctio
n
-
b
ase
d
d
o
m
ain
s
,
it
is
r
elea
s
ed
u
n
d
er
th
e
Ap
ac
h
e
2
.
0
licen
s
e.
As
a
s
ig
n
if
ican
t
a
d
v
an
ce
m
en
t
in
lan
g
u
ag
e
m
o
d
el
s
,
‘
Falco
n
-
7B
-
I
n
s
tr
u
ct
’
s
er
v
es
as
a
p
o
wer
f
u
l
an
d
o
p
en
ly
licen
s
ed
c
o
n
tr
ib
u
tio
n
t
o
th
e
f
ield
[
7
]
.
T
h
e
‘
Mistra
l
-
7B
-
v
0
.
1
’
LLM
is
a
7
-
b
illi
o
n
-
p
ar
am
eter
g
en
e
r
ativ
e
tex
t
m
o
d
el
th
at
o
u
t
p
er
f
o
r
m
s
‘
L
lam
a
-
2
-
1
3
B
’
ac
r
o
s
s
all
ass
es
s
ed
b
en
ch
m
ar
k
s
.
Desig
n
ed
as
a
tr
an
s
f
o
r
m
er
-
b
ased
ar
c
h
itectu
r
e,
it
in
teg
r
ates
ad
v
an
ce
d
f
ea
tu
r
es
s
u
ch
as
Gr
o
u
p
ed
-
Qu
er
y
Atten
tio
n
,
Sli
d
in
g
-
W
in
d
o
w
Atten
tio
n
,
an
d
a
B
y
te
-
f
allb
ac
k
B
PE
to
k
en
izer
.
No
tab
ly
,
‘
Mistra
l
7
B
’
f
u
n
ctio
n
s
as
a
b
ase
m
o
d
el
an
d
d
o
es
n
o
t
in
clu
d
e
m
o
d
e
r
atio
n
m
ec
h
an
is
m
s
,
as
it
is
p
u
r
ely
a
p
r
e
-
tr
ain
e
d
m
o
d
el
[
8
]
.
T
h
e
'
L
lam
a
-
v2
-
1
3
B
'
m
o
d
el
is
a
ca
r
ef
u
lly
f
in
e
-
tu
n
ed
lan
g
u
ag
e
m
o
d
el
tailo
r
ed
f
o
r
d
ial
o
g
u
e
-
b
ased
ap
p
licatio
n
s
an
d
co
m
m
e
r
cial
u
s
e.
B
u
ilt
o
n
an
o
p
tim
ized
tr
an
s
f
o
r
m
er
ar
c
h
itectu
r
e,
‘
L
lam
a
2
’
f
u
n
ctio
n
s
as
an
au
to
-
r
eg
r
ess
iv
e
m
o
d
el.
I
t
e
m
p
lo
y
s
a
d
u
al
o
p
tim
izatio
n
s
tr
ateg
y
co
m
b
in
in
g
s
u
p
er
v
is
ed
f
i
n
e
-
tu
n
in
g
(
SF
T
)
an
d
r
ein
f
o
r
ce
m
e
n
t
lear
n
in
g
with
h
u
m
an
f
ee
d
b
ac
k
(
R
L
HF)
to
alig
n
with
h
u
m
an
p
r
ef
e
r
en
ce
s
,
en
s
u
r
in
g
b
o
th
u
s
ef
u
ln
ess
an
d
s
af
ety
[
9
]
.
Me
an
wh
ile,
‘
C
o
d
eL
lam
a
3
4
B
v
2
,
’
r
ef
in
ed
f
r
o
m
‘
Ph
in
d
-
C
o
d
eL
lam
a
-
34B
-
v
1
,
’
ac
h
iev
es a
n
im
p
r
ess
iv
e
7
3
.
8
%
p
ass
@
1
o
n
Hu
m
an
E
v
al,
estab
lis
h
in
g
its
elf
as th
e
lead
in
g
o
p
en
-
s
o
u
r
ce
m
o
d
el
in
its
d
o
m
ain
[
1
0
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
446
-
4
5
4
448
‘
My
th
o
Ma
x
1
3
B
’
,
as
cited
in
[
1
1
]
,
r
ep
r
esen
ts
a
p
in
n
ac
le
in
f
in
e
-
tu
n
e
d
lan
g
u
ag
e
m
o
d
els,
o
r
ig
in
atin
g
f
r
o
m
th
e
r
o
b
u
s
t
‘
L
lam
a
2
1
3
B
’
.
An
e
v
o
lu
tio
n
o
f
Gr
y
p
h
e'
s
‘
My
th
o
Ma
x
L
2
1
3
B
’
m
o
d
e
l
ca
r
d
,
th
is
v
a
r
ian
t
in
tr
o
d
u
ce
s
a
r
ef
in
ed
f
u
s
io
n
tech
n
iq
u
e,
m
e
r
g
in
g
My
t
h
o
L
o
g
ic
-
L
2
an
d
Hu
g
in
n
th
r
o
u
g
h
an
ex
p
er
im
en
tal
ten
s
o
r
ty
p
e
m
er
g
e.
W
h
at
s
ets
‘
My
th
o
Ma
x
1
3
B
’
ap
a
r
t
is
its
e
m
p
h
asis
o
n
e
n
r
ich
e
d
d
escr
ip
tio
n
s
an
d
r
o
lep
la
y
ca
p
ab
ilit
ies,
m
ak
in
g
it
a
g
o
-
t
o
ch
o
o
s
e
f
o
r
n
ar
r
ativ
e
task
s
.
No
tab
ly
,
th
e
m
o
d
el
em
p
lo
y
s
Alp
ac
a
f
o
r
m
attin
g
,
en
s
u
r
in
g
a
v
is
u
ally
co
n
s
is
ten
t
an
d
en
g
ag
in
g
o
u
t
p
u
t.
T
h
e
in
n
o
v
ativ
e
ap
p
r
o
ac
h
o
f
allo
win
g
m
o
r
e
in
ter
m
i
n
g
lin
g
o
f
Hu
g
in
n
with
th
e
m
o
d
el'
s
ten
s
o
r
s
en
h
an
ce
s
o
v
er
all
co
h
er
en
ce
.
I
n
ess
en
ce
,
‘
My
th
o
M
ax
1
3
B
’
co
m
b
in
es
f
an
tasy
elem
en
ts
with
s
tr
u
ctu
r
al
f
in
ess
e,
o
f
f
e
r
in
g
a
p
o
wer
f
u
l
to
o
l
f
o
r
im
m
e
r
s
iv
e
s
to
r
y
telli
n
g
an
d
r
o
lep
lay
in
g
ex
p
er
ien
ce
s
.
T
h
e
‘
to
p
p
y
-
m
-
7
b
’
m
o
d
el
[
1
2
]
,
ac
ce
s
s
ib
le
at
‘
u
n
d
i9
5
/to
p
p
y
-
m
-
7
b
’
,
b
o
asti
n
g
an
im
p
r
ess
iv
e
7
b
illi
o
n
p
ar
am
eter
s
,
r
ep
r
esen
ts
a
co
n
v
er
g
en
ce
o
f
in
f
lu
en
tial
m
o
d
e
ls
f
ac
ilit
ated
b
y
th
e
in
n
o
v
ativ
e
‘
task
_
ar
ith
m
etic’
m
er
g
e
m
eth
o
d
f
r
o
m
‘
m
er
g
ek
it’.
Me
r
g
e
d
m
o
d
els
in
clu
d
e
‘
No
u
s
R
esear
ch
/No
u
s
-
C
ap
y
b
ar
a
-
7B
-
V1
.
9
’
,
‘
Hu
g
g
in
g
Face
H4
/zep
h
y
r
-
7b
-
b
eta’
,
‘
lem
o
n
ilia
/As
h
h
L
im
a
R
P
-
Mis
tr
al
-
7
B
’
,
‘
Vu
lk
an
e/
1
2
0
-
Day
s
-
of
-
So
d
o
m
-
L
o
R
A
-
Mistra
l
-
7
b
’
,
an
d
‘
Un
d
i
9
5
/Mist
r
al
-
p
ip
p
a
-
s
h
ar
e
g
p
t
-
7b
-
q
lo
r
a’
.
T
h
is
co
llab
o
r
ativ
e
ef
f
o
r
t
y
ield
s
a
p
o
wer
f
u
l
m
o
d
el,
d
e
m
o
n
s
tr
atin
g
t
h
e
f
o
r
e
f
r
o
n
t o
f
r
esear
ch
an
d
i
n
n
o
v
atio
n
in
th
e
f
ield
o
f
NL
P.
‘
My
th
o
Mist
7
B
’
,
av
ailab
le
at
‘
g
r
y
p
h
e/m
y
th
o
m
is
t
-
7
b
’
[
1
3
]
,
is
a
s
o
p
h
is
ticated
ch
at
-
b
ased
lan
g
u
ag
e
m
o
d
el
d
esig
n
e
d
to
en
h
an
ce
r
o
lep
lay
in
g
ex
p
er
ien
ce
s
.
W
ith
a
n
ex
p
a
n
s
iv
e
co
n
tex
t
o
f
3
2
,
7
6
8
to
k
en
s
,
it
o
f
f
er
s
a
s
ea
m
less
co
n
v
er
s
atio
n
al
f
lo
w.
C
r
ea
ted
b
y
th
e
m
aster
m
in
d
b
eh
in
d
‘
M
y
th
o
Ma
x
’
,
th
is
m
o
d
e
l
s
k
illfu
lly
m
er
g
es
s
ev
er
al
p
r
o
m
in
e
n
t
m
o
d
els,
in
clu
d
in
g
‘
Neu
r
al
C
h
at
7
B
’
,
‘
A
ir
o
b
o
r
o
s
7
b
’
,
‘
T
o
p
p
y
M
7
B
’
,
‘
Z
ep
h
y
r
7
b
b
eta’
,
‘
No
u
s
C
ap
y
b
a
r
a
3
4
B
’
,
‘
Op
en
Her
em
es
2
.
5
’
,
a
n
d
m
o
r
e
.
T
h
e
in
teg
r
atio
n
aim
s
t
o
m
i
n
im
ize
wo
r
d
an
ticip
atio
n
,
r
ef
in
e
m
in
is
tr
atio
n
s
,
an
d
m
iti
g
ate
th
e
p
r
esen
ce
o
f
u
n
d
esir
ab
le
wo
r
d
s
,
p
r
o
v
id
in
g
a
n
en
r
ich
ed
an
d
tailo
r
ed
r
o
lep
lay
in
g
en
v
ir
o
n
m
e
n
t.
As
cited
b
y
A
lp
in
d
ale
[
1
4
]
,
‘
Go
liath
1
2
0
B
’
,
is
a
f
o
r
m
id
a
b
le
lan
g
u
a
g
e
m
o
d
el
th
at
lev
er
ag
es
an
ex
ten
s
iv
e
co
n
tex
t
o
f
6
,
1
4
4
t
o
k
en
s
to
p
r
o
v
id
e
a
r
ich
an
d
n
u
an
ce
d
ch
at
ex
p
er
ien
ce
.
C
r
ea
ted
th
r
o
u
g
h
th
e
am
alg
am
atio
n
o
f
two
f
in
ely
-
t
u
n
ed
L
lam
a
7
0
B
m
o
d
els,
th
is
lar
g
e
L
L
M
b
o
asts
an
im
p
r
ess
iv
e
p
ar
am
ete
r
co
u
n
t
o
f
1
2
0
b
illi
o
n
.
T
h
e
m
o
d
el
s
ea
m
less
ly
in
teg
r
ates
th
e
ca
p
a
b
ilit
ies
o
f
Xwin
a
n
d
E
u
r
y
ale,
r
es
u
ltin
g
in
a
p
o
wer
f
u
l
an
d
v
er
s
atile
lan
g
u
ag
e
g
e
n
er
at
io
n
m
o
d
el.
‘
PaL
M
2
’
,
tailo
r
ed
f
o
r
ch
atb
o
t
in
ter
ac
tio
n
s
,
ex
ce
ls
in
ass
is
ti
n
g
with
in
q
u
ir
ies
r
elate
d
to
c
o
d
in
g
.
As
a
cu
ttin
g
-
ed
g
e
la
n
g
u
a
g
e
m
o
d
el
,
‘
PaL
M
2
’
b
o
asts
en
h
an
ce
d
m
u
ltil
in
g
u
al
p
r
o
f
icien
cy
,
a
d
v
an
ce
d
r
ea
s
o
n
in
g
ab
ilit
ies,
an
d
an
ad
e
p
t
u
n
d
er
s
tan
d
in
g
o
f
co
d
in
g
co
n
ce
p
ts
.
Dev
elo
p
ed
b
y
Go
o
g
le,
‘
PaL
M
2
’
is
a
ch
at
-
b
is
o
n
m
o
d
el
d
esig
n
ed
t
o
h
a
n
d
le
c
o
d
e
-
r
elate
d
q
u
esti
o
n
s
e
f
f
ec
ti
v
ely
.
No
ta
b
ly
,
it
s
u
p
p
o
r
ts
a
s
u
b
s
tan
tial
co
n
tex
t
win
d
o
w
o
f
8
,
0
0
0
to
k
en
s
,
f
ac
ili
tatin
g
co
m
p
r
e
h
en
s
iv
e
an
d
co
n
tex
tu
ally
r
ich
c
o
n
v
e
r
s
atio
n
s
[
1
5
]
.
An
th
r
o
p
ic
i
n
tr
o
d
u
ce
d
C
lau
d
e
m
o
d
els,
as
cited
i
n
[
1
6
]
,
[
1
7
]
,
o
f
ten
wo
r
k
well
f
o
r
wr
iti
n
g
,
ed
itin
g
,
s
u
m
m
ar
izin
g
,
s
ea
r
ch
in
g
,
an
d
g
en
er
al,
o
p
e
n
-
en
d
ed
co
n
v
er
s
atio
n
s
.
C
o
n
s
titu
tio
n
al
ar
tific
ial
in
tellig
en
ce
(
AI
)
an
d
u
n
s
u
p
er
v
is
ed
lear
n
in
g
ar
e
u
s
ed
in
th
e
tr
ain
in
g
o
f
C
lau
d
e
m
o
d
els,
wh
ich
ar
e
g
en
e
r
al
-
p
u
r
p
o
s
e
L
L
Ms
th
at
em
p
lo
y
tr
a
n
s
f
o
r
m
er
ar
ch
itec
tu
r
e.
C
lau
d
e
m
o
d
els
ar
e
co
r
p
o
r
ate
ap
p
licatio
n
-
s
p
ec
if
ic
m
o
d
els.
T
h
e
ch
at
co
m
p
letio
n
m
o
d
el,
C
lau
d
e
v
1
,
is
p
er
f
e
ct
f
o
r
co
n
d
en
s
in
g
,
ex
am
in
in
g
,
a
n
d
s
ea
r
ch
i
n
g
len
g
th
y
te
x
ts
an
d
d
is
cu
s
s
io
n
s
in
o
r
d
er
to
g
ain
a
s
o
p
h
is
ticated
g
r
asp
o
f
in
tr
i
ca
te
s
u
b
jects
an
d
th
eir
co
n
n
ec
tio
n
s
th
r
o
u
g
h
o
u
t
ex
tr
em
ely
lo
n
g
te
x
t
s
eg
m
en
ts
.
T
h
e
f
lag
s
h
ip
m
o
d
el,
‘
C
lau
d
e
v
2
.
0
’
,
h
as
lo
n
g
er
r
ef
lex
es
an
d
p
er
f
o
r
m
s
b
etter
.
I
t
h
as
an
asto
u
n
d
in
g
1
0
0
k
to
k
en
ca
p
ac
ity
f
o
r
a
co
n
tex
t
win
d
o
w.
T
h
e
a
d
v
an
ce
d
L
L
M
‘
C
lau
d
e
v
2
.
1
’
h
as
a
2
0
0
K
to
k
en
co
n
tex
t w
in
d
o
w
an
d
a
2
x
r
ed
u
ctio
n
in
to
k
en
u
s
ag
e.
A
s
r
e
f
e
r
e
n
c
e
d
i
n
[
1
8
]
,
‘
N
o
u
s
-
H
e
r
m
e
s
-
L
l
a
m
a
2
-
1
3
B
’
i
s
a
n
a
d
v
a
n
c
e
d
l
a
n
g
u
a
g
e
m
o
d
e
l
m
e
t
i
c
u
l
o
u
s
l
y
f
i
n
e
-
t
u
n
e
d
o
n
a
l
a
r
g
e
d
a
t
a
s
e
t
o
f
o
v
e
r
3
0
0
,
0
0
0
i
n
s
t
r
u
c
t
i
o
n
s
.
I
t
i
s
d
i
s
t
i
n
g
u
i
s
h
e
d
b
y
i
t
s
a
b
i
l
i
t
y
t
o
g
e
n
e
r
a
t
e
e
x
t
e
n
d
e
d
r
e
s
p
o
n
s
e
s
,
m
i
n
i
m
i
z
e
h
a
l
l
u
c
i
n
a
t
i
o
n
s
,
a
n
d
o
p
e
r
a
t
e
w
i
t
h
o
u
t
O
p
e
n
A
I
c
e
n
s
o
r
s
h
i
p
m
e
c
h
a
n
i
s
m
s
i
n
i
t
s
s
y
n
t
h
e
t
i
c
t
r
a
i
n
i
n
g
d
a
t
a
.
P
r
i
m
a
r
i
l
y
t
r
a
i
n
e
d
o
n
s
y
n
t
h
e
t
i
c
G
P
T
-
4
o
u
t
p
u
t
s
,
t
h
e
m
o
d
e
l
b
e
n
e
f
i
t
s
f
r
o
m
h
i
g
h
-
q
u
a
l
i
t
y
G
P
T
-
4
d
a
t
a
s
e
t
s
,
e
n
h
a
n
c
i
n
g
i
t
s
p
r
o
f
i
c
i
e
n
c
y
i
n
k
n
o
w
l
e
d
g
e
d
e
l
i
v
e
r
y
,
t
a
s
k
e
x
e
c
u
t
i
o
n
,
a
n
d
s
t
y
l
i
s
t
i
c
g
e
n
e
r
a
t
i
o
n
.
T
h
e
PP
L
X
m
o
d
els,
ex
em
p
li
f
ied
b
y
‘
PP
L
X
-
7B
-
On
lin
e’
an
d
‘
PP
L
X
-
70B
-
On
lin
e’
,
r
e
d
ef
in
e
th
e
lan
d
s
ca
p
e
o
f
L
L
Ms
b
y
s
p
ec
if
ically
tack
lin
g
two
p
r
ev
alen
t
ch
allen
g
es.
Un
lik
e
m
an
y
L
L
Ms,
th
ese
m
o
d
els
p
r
io
r
itize
d
eliv
e
r
in
g
r
esp
o
n
s
e
s
th
at
ar
e
n
o
t
o
n
ly
h
elp
f
u
l
a
n
d
f
ac
t
u
al
b
u
t
also
u
p
-
to
-
d
ate,
o
v
e
r
co
m
in
g
th
e
lim
itatio
n
o
f
o
u
td
ated
in
f
o
r
m
a
tio
n
.
Ad
d
itio
n
ally
,
th
e
y
ad
d
r
es
s
in
ac
cu
r
ac
ies
co
m
m
o
n
ly
ass
o
ciate
d
with
L
L
Ms,
m
in
im
izin
g
h
allu
cin
atio
n
s
a
n
d
en
s
u
r
i
n
g
r
esp
o
n
s
es
ar
e
ac
c
u
r
ate
an
d
r
eliab
le.
T
h
r
o
u
g
h
t
h
ese
ad
v
an
ce
m
en
ts
,
PP
L
X
m
o
d
els
s
et
a
n
ew
s
ta
n
d
ar
d
b
y
p
r
o
v
id
in
g
a
u
n
iq
u
e
b
len
d
o
f
r
ea
l
-
tim
e
r
elev
an
ce
,
p
r
ec
is
io
n
,
an
d
h
elp
f
u
ln
ess
in
th
eir
la
n
g
u
a
g
e
g
en
er
atio
n
ca
p
a
b
ilit
ies [
1
9
]
,
[
2
0
]
.
T
h
is
p
ap
er
o
f
f
e
r
s
an
in
-
d
ep
t
h
ex
p
l
o
r
atio
n
o
f
te
x
t
s
u
m
m
a
r
izatio
n
,
an
al
y
zin
g
a
d
iv
e
r
s
e
s
et
o
f
2
5
L
L
Ms,
f
r
o
m
co
n
v
en
tio
n
al
m
o
d
els
to
th
e
latest
in
n
o
v
atio
n
s
.
I
t
th
o
r
o
u
g
h
ly
ex
am
in
es
t
h
e
ca
p
ab
ilit
ies
an
d
lim
itatio
n
s
o
f
s
tate
-
of
-
th
e
-
ar
t
L
L
Ms
b
y
ex
p
er
im
en
tin
g
w
ith
v
ar
io
u
s
h
y
p
er
p
ar
am
eter
s
an
d
ass
ess
in
g
th
e
g
en
er
ated
s
u
m
m
ar
ies
u
s
in
g
estab
lis
h
ed
m
etr
ics
s
u
ch
as
b
ilin
g
u
al
ev
alu
atio
n
u
n
d
er
s
tu
d
y
(
B
L
E
U
)
s
co
r
e,
r
ec
all
-
o
r
ien
ted
u
n
d
e
r
s
tu
d
y
f
o
r
g
is
tin
g
ev
alu
atio
n
(
R
OUGE
)
s
co
r
e,
an
d
B
E
R
T
s
co
r
e.
As
a
v
alu
a
b
le
r
ef
er
e
n
ce
,
th
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
I
n
ve
s
tig
a
tin
g
th
e
r
ec
a
ll e
fficie
n
cy
in
a
b
s
tr
a
ctive
s
u
mma
r
iz
a
t
io
n
:
a
n
ex
p
erimen
ta
l
…
(
S
u
r
a
b
h
i A
n
u
r
a
d
h
a
)
449
s
tu
d
y
p
r
o
v
id
es
ess
en
tial
in
s
ig
h
ts
f
o
r
th
o
s
e
s
ee
k
in
g
to
h
ar
n
e
s
s
L
L
Ms
in
NL
P
ap
p
licatio
n
s
an
d
p
a
v
es
th
e
wa
y
f
o
r
d
e
v
elo
p
in
g
ad
v
an
ce
d
g
en
e
r
ativ
e
AI
s
o
lu
tio
n
s
to
tack
le
a
wid
e
r
an
g
e
o
f
b
u
s
in
ess
ch
allen
g
es.
B
a
s
y
a
l
a
n
d
S
a
n
g
h
v
i
[
2
]
i
n
t
h
e
i
r
s
t
u
d
y
u
s
e
d
t
h
r
e
e
LLM
s
M
P
T
-
7b
-
i
n
s
t
r
u
c
t
,
F
a
l
c
o
n
-
7b
-
i
n
s
t
r
u
c
t
,
a
n
d
O
p
e
n
A
I
C
h
a
t
-
G
P
T
w
i
t
h
t
w
o
d
a
t
a
s
e
t
s
C
N
N
D
a
i
l
y
M
a
i
l
a
n
d
X
S
u
m
.
T
h
e
i
r
f
o
c
u
s
i
s
o
n
e
x
t
r
a
c
t
i
v
e
s
u
m
m
a
r
i
z
a
t
i
o
n
o
f
n
e
w
s
a
r
t
i
c
l
e
s
,
p
r
i
m
a
r
i
l
y
e
v
a
l
u
a
t
i
n
g
p
e
r
f
o
r
m
a
n
c
e
u
s
i
n
g
p
r
e
c
i
s
i
o
n
.
E
x
t
r
a
c
t
i
v
e
s
u
m
m
a
r
i
e
s
,
h
o
w
e
v
e
r
,
o
f
t
e
n
r
e
s
u
l
t
i
n
a
d
i
s
j
o
i
n
t
e
d
s
e
q
u
e
n
c
e
o
f
s
e
n
t
e
n
c
e
s
d
u
e
t
o
t
h
e
s
i
m
p
l
e
c
o
n
c
a
t
e
n
a
t
i
o
n
o
f
s
e
n
t
e
n
c
e
s
f
r
o
m
v
a
r
i
o
u
s
p
a
r
t
s
o
f
t
h
e
t
e
x
t
,
l
e
a
d
i
n
g
t
o
a
l
a
c
k
o
f
n
a
t
u
r
a
l
f
l
o
w
a
n
d
c
o
h
e
r
e
n
c
e
.
A
d
d
i
t
i
o
n
a
l
l
y
,
b
e
c
a
u
s
e
e
x
t
r
a
c
t
i
v
e
s
u
m
m
a
r
i
z
a
t
i
o
n
c
a
n
n
o
t
r
e
p
h
r
a
s
e
o
r
c
o
n
d
e
n
s
e
i
n
f
o
r
m
a
t
i
o
n
a
n
d
r
e
l
i
e
s
s
o
l
e
l
y
o
n
t
h
e
o
r
i
g
i
n
a
l
w
o
r
d
i
n
g
,
i
t
f
a
i
l
s
t
o
c
o
n
v
e
y
k
e
y
i
d
e
a
s
c
o
n
c
i
s
e
l
y
.
I
n
c
o
n
t
r
a
s
t
,
o
u
r
r
e
s
e
a
r
c
h
f
o
c
u
s
e
s
o
n
g
e
n
e
r
a
t
i
n
g
a
b
s
t
r
a
c
t
i
v
e
s
u
m
m
a
r
i
e
s
f
o
r
s
c
i
e
n
t
i
f
i
c
s
c
h
o
l
a
r
l
y
d
o
c
u
m
e
n
t
s
,
i
n
c
o
r
p
o
r
a
t
i
n
g
a
d
i
v
e
r
s
e
s
e
t
o
f
2
5
a
d
v
a
n
c
e
d
L
L
M
s
.
O
u
r
p
r
i
m
a
r
y
f
o
c
u
s
i
s
o
n
r
e
c
a
l
l
-
o
r
i
e
n
t
e
d
e
v
a
l
u
a
t
i
o
n
t
o
e
n
s
u
r
e
c
o
m
p
r
e
h
e
n
s
i
v
e
i
n
c
l
u
s
i
o
n
o
f
r
e
l
e
v
a
n
t
d
e
t
a
i
l
s
.
T
h
e
k
e
y
o
b
j
e
c
t
i
v
e
s
o
f
o
u
r
r
e
s
e
a
r
c
h
w
o
r
k
a
r
e
:
−
T
o
im
p
r
o
v
e
n
atu
r
al
f
lo
w
an
d
co
h
er
en
ce
in
s
u
m
m
a
r
ies
.
Ou
r
r
esear
ch
aim
s
to
ad
d
r
ess
th
e
is
s
u
e
o
f
ex
tr
ac
tiv
e
s
u
m
m
ar
ies
o
f
ten
r
esu
ltin
g
in
a
d
is
jo
in
ted
s
eq
u
en
ce
o
f
s
en
ten
ce
s
.
B
y
m
o
v
in
g
b
ey
o
n
d
s
im
p
l
y
co
n
ca
ten
atin
g
s
en
ten
ce
s
f
r
o
m
d
if
f
er
e
n
t
p
ar
t
s
o
f
th
e
tex
t,
we
s
ee
k
to
en
h
an
ce
th
e
n
atu
r
al
f
lo
w
an
d
c
o
h
er
en
ce
o
f
th
e
g
en
er
ated
s
u
m
m
ar
ies.
Ad
d
itio
n
ally
,
o
u
r
g
o
al
is
to
o
v
er
c
o
m
e
th
e
lim
itatio
n
s
o
f
ex
tr
ac
tiv
e
s
u
m
m
ar
izatio
n
,
wh
ich
ca
n
n
o
t
r
ep
h
r
ase
o
r
co
n
d
en
s
e
in
f
o
r
m
atio
n
an
d
r
elies
s
o
lely
o
n
th
e
o
r
ig
in
al
w
o
r
d
in
g
,
o
f
ten
f
ailin
g
to
co
n
v
ey
k
ey
i
d
ea
s
co
n
cisely
.
−
T
o
en
s
u
r
e
co
m
p
r
e
h
en
s
iv
e
in
c
lu
s
io
n
o
f
r
elev
a
n
t
d
etails.
R
ec
o
g
n
izin
g
th
at
s
cien
tific
d
o
c
u
m
en
ts
co
n
tain
d
en
s
e,
cr
itical
in
f
o
r
m
atio
n
ess
en
tial
f
o
r
u
n
d
e
r
s
tan
d
in
g
r
esear
ch
,
o
u
r
r
esear
ch
f
o
cu
s
es
o
n
ac
h
iev
in
g
h
i
g
h
r
ec
all.
T
h
is
en
s
u
r
es
th
at
th
e
s
u
m
m
ar
ies
in
clu
d
e
as
m
an
y
r
el
ev
an
t
d
etails
an
d
k
e
y
p
o
i
n
ts
f
r
o
m
th
e
o
r
i
g
in
al
d
o
cu
m
e
n
t
as
p
o
s
s
ib
le,
th
er
e
b
y
p
r
o
v
id
in
g
co
m
p
r
eh
e
n
s
iv
e
an
d
ac
cu
r
ate
r
ep
r
esen
tatio
n
s
o
f
th
e
s
o
u
r
ce
m
ater
ial.
T
h
e
r
em
ai
n
d
er
o
f
th
e
p
ap
e
r
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws:
s
ec
tio
n
2
i
n
tr
o
d
u
ce
s
th
e
d
ataset
an
d
ev
alu
atio
n
m
etr
ics
u
s
ed
to
m
ea
s
u
r
e
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
Sectio
n
3
d
etails
th
e
ex
p
er
im
en
tal
s
etu
p
,
wh
ile
s
ec
tio
n
4
p
r
esen
ts
th
e
i
n
f
er
en
ce
r
esu
lts
o
f
v
ar
io
u
s
L
L
Ms.
L
astl
y
,
s
ec
tio
n
5
co
n
clu
d
es
th
e
s
tu
d
y
a
n
d
o
u
tlin
es
d
ir
ec
tio
n
s
f
o
r
f
u
tu
r
e
e
n
h
an
ce
m
e
n
ts
.
2.
DATAS
E
T
AND
E
VAL
UAT
I
O
N
M
E
T
R
I
CS
I
n
th
is
s
tu
d
y
,
we
co
n
d
u
cted
ex
p
er
im
en
ts
an
d
ev
alu
atio
n
s
u
s
in
g
th
e
SciTL
DR
d
ataset,
o
r
ig
in
ally
in
tr
o
d
u
ce
d
b
y
C
ac
h
o
la
et
a
l.
[
2
1
]
i
n
"
T
L
DR
:
ex
tr
em
e
s
u
m
m
ar
izatio
n
o
f
s
cien
tific
d
o
cu
m
e
n
ts
"
.
T
h
e
SciTL
DR
d
ataset
is
d
esig
n
ed
to
f
ac
ilit
ate
ex
tr
em
e
s
u
m
m
ar
izatio
n
task
s
in
th
e
co
n
tex
t
o
f
s
cien
tific
d
o
c
u
m
en
ts
.
Ou
r
s
tu
d
y
aim
ed
to
ass
es
s
th
e
p
er
f
o
r
m
a
n
ce
o
f
s
p
ec
if
ic
m
o
d
els
in
ex
tr
em
e
s
u
m
m
ar
izatio
n
,
lev
er
a
g
i
n
g
th
is
d
ataset
as
a
b
en
ch
m
ar
k
.
W
e
f
o
llo
wed
a
m
eth
o
d
o
lo
g
y
th
at
in
v
o
lv
ed
im
p
lem
en
tin
g
v
ar
io
u
s
al
g
o
r
ith
m
s
an
d
ass
ess
in
g
th
eir
s
u
m
m
ar
izatio
n
ca
p
a
b
ilit
ies,
u
s
in
g
m
etr
ics
r
elev
an
t
to
ex
tr
em
e
s
u
m
m
ar
izatio
n
task
s
.
T
h
e
ch
o
ice
o
f
th
e
SciTL
DR
d
ataset
p
r
o
v
id
ed
a
s
tan
d
ar
d
ized
an
d
ch
allen
g
in
g
s
et
o
f
s
cien
tific
d
o
cu
m
e
n
t
s
,
allo
win
g
f
o
r
a
co
m
p
r
eh
e
n
s
iv
e
ev
alu
atio
n
o
f
t
h
e
s
u
m
m
ar
izatio
n
m
o
d
els.
2
.
1
.
Da
t
a
s
et
SciTL
DR
:
t
h
i
s
d
ataset,
co
m
p
r
is
in
g
m
u
ltip
le
tar
g
ets,
en
co
m
p
ass
es
5
.
4
K
T
L
DR
s
ex
tr
ac
ted
f
r
o
m
3
.
2
K
p
u
b
licatio
n
s
.
T
h
e
d
ataset
in
c
o
r
p
o
r
ates
b
o
t
h
T
L
DR
s
wr
itte
n
b
y
au
th
o
r
s
an
d
th
o
s
e
d
er
i
v
ed
b
y
ex
p
e
r
ts
.
T
h
e
ex
p
er
t
-
d
e
r
iv
ed
s
u
m
m
ar
ies
a
r
e
o
b
tain
e
d
t
h
r
o
u
g
h
a
d
is
tin
ctiv
e
an
n
o
tatio
n
p
r
o
ce
s
s
d
esig
n
ed
to
r
ed
u
ce
th
e
an
n
o
tatio
n
w
o
r
k
lo
a
d
wh
ile
en
s
u
r
in
g
th
e
p
r
o
d
u
ctio
n
o
f
h
i
g
h
-
q
u
ality
s
u
m
m
ar
ies.
2
.
2
.
E
v
a
lua
t
i
o
n
m
et
rics
T
o
ass
ess
th
e
ef
f
ec
tiv
en
ess
an
d
ac
cu
r
ac
y
o
f
s
u
m
m
a
r
ies
g
en
e
r
ated
b
y
v
a
r
io
u
s
LLMs
,
we
u
ti
lized
a
s
et
o
f
well
-
estab
lis
h
ed
ev
alu
atio
n
m
etr
ics.
T
h
e
B
L
E
U
s
co
r
e
is
a
wid
ely
u
s
ed
m
etr
ic
f
o
r
e
v
alu
atin
g
m
ac
h
i
n
e
-
g
en
er
ated
te
x
t
ac
r
o
s
s
d
if
f
e
r
e
n
t
NL
P
task
s
,
in
clu
d
in
g
te
x
t
s
u
m
m
ar
izatio
n
[
2
2
]
.
I
t
m
ea
s
u
r
es
h
o
w
clo
s
ely
a
g
en
er
ated
s
u
m
m
ar
y
alig
n
s
with
o
n
e
o
r
m
o
r
e
r
ef
e
r
en
ce
s
u
m
m
ar
ies,
p
r
o
v
id
in
g
a
q
u
an
tit
ativ
e
ass
ess
m
en
t
o
f
p
r
ec
is
io
n
an
d
te
x
tu
al
o
v
e
r
la
p
.
T
h
e
B
L
E
U
s
co
r
e
is
ca
lcu
lated
b
y
c
o
m
p
ar
i
n
g
n
-
g
r
a
m
s
(
s
eq
u
en
ce
s
o
f
co
n
s
ec
u
tiv
e
wo
r
d
s
o
r
to
k
en
s
)
in
th
e
g
en
er
ated
s
u
m
m
ar
y
t
o
th
o
s
e
in
th
e
r
ef
er
en
ce
s
u
m
m
ar
ies.
Pre
cisi
o
n
is
d
eter
m
in
ed
b
y
t
h
e
p
r
o
p
o
r
tio
n
o
f
m
atch
in
g
n
-
g
r
am
s
,
wh
i
le
a
b
r
ev
ity
p
en
alty
is
ap
p
li
ed
to
p
r
ev
en
t
t
h
e
o
v
er
v
alu
atio
n
o
f
ex
ce
s
s
iv
ely
s
h
o
r
t
s
u
m
m
ar
ies.
A
h
ig
h
e
r
B
L
E
U
s
co
r
e,
r
an
g
in
g
f
r
o
m
0
to
1
,
in
d
icate
s
a
s
tr
o
n
g
er
c
o
r
r
esp
o
n
d
en
ce
b
et
wee
n
th
e
g
e
n
er
ated
an
d
r
ef
er
en
ce
s
u
m
m
ar
ies,
r
ef
lectin
g
im
p
r
o
v
ed
c
o
n
ten
t
ac
cu
r
ac
y
an
d
s
tr
u
ctu
r
al
c
o
h
er
e
n
ce
.
T
h
e
R
OUGE
s
co
r
e
ev
alu
ates
th
e
s
im
ilar
ity
b
etwe
en
a
g
en
er
a
ted
tex
t
an
d
o
n
e
o
r
m
o
r
e
r
ef
er
en
ce
tex
ts
b
y
a
n
aly
zin
g
th
e
o
v
e
r
lap
o
f
n
-
g
r
am
s
an
d
w
o
r
d
s
eq
u
en
c
es
[
2
3
]
.
I
t
co
m
p
r
is
es
s
ev
er
al
m
etr
ics,
in
clu
d
in
g
R
OUGE
-
N
(
wh
ich
c
o
n
s
id
er
s
u
n
ig
r
a
m
s
an
d
b
ig
r
am
s
)
,
R
OUGE
-
L
(
wh
ich
f
o
cu
s
es
o
n
th
e
lo
n
g
est
co
m
m
o
n
s
u
b
s
eq
u
en
ce
)
,
an
d
R
OUGE
-
W
(
wh
ich
m
ea
s
u
r
es
wo
r
d
o
v
er
lap
)
.
A
h
ig
h
er
R
OUGE
s
co
r
e,
ty
p
ically
r
a
n
g
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
446
-
4
5
4
450
f
r
o
m
0
to
1
,
s
ig
n
if
ies
a
g
r
ea
te
r
alig
n
m
e
n
t
b
etwe
en
th
e
g
e
n
e
r
ated
an
d
r
ef
er
en
ce
s
u
m
m
a
r
ies,
o
f
f
er
i
n
g
v
alu
ab
le
in
s
ig
h
ts
in
to
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
s
u
m
m
a
r
izatio
n
m
o
d
el
[
2
4
]
.
B
E
R
T
s
co
r
e
u
tili
ze
s
co
n
tex
t
u
al
em
b
e
d
d
in
g
s
f
r
o
m
th
e
B
E
R
T
m
o
d
el
t
o
m
ea
s
u
r
e
th
e
s
im
ilar
ity
b
etwe
en
a
g
en
er
ate
d
s
u
m
m
a
r
y
an
d
its
r
ef
er
e
n
ce
s
u
m
m
a
r
ies.
Desig
n
ed
to
ca
p
tu
r
e
th
e
n
u
a
n
c
es
o
f
lan
g
u
ag
e
an
d
co
n
tex
t,
t
h
is
m
etr
ic
p
r
o
v
i
d
es
a
p
o
wer
f
u
l
a
p
p
r
o
ac
h
f
o
r
ass
ess
in
g
th
e
q
u
ality
an
d
r
ele
v
an
ce
o
f
th
e
g
en
er
ate
d
co
n
ten
t
[
2
5
]
,
[
2
6
]
.
B
y
co
m
p
u
tin
g
th
ese
m
etr
ics
f
o
r
s
u
m
m
a
r
ies
g
en
er
ated
b
y
v
ar
io
u
s
L
L
Ms,
o
u
r
g
o
al
is
to
p
r
o
v
id
e
a
co
m
p
r
eh
e
n
s
iv
e
a
s
s
es
s
m
en
t
o
f
th
eir
p
er
f
o
r
m
an
ce
.
T
h
is
ev
al
u
atio
n
e
q
u
i
p
s
r
esear
ch
er
s
an
d
p
r
ac
titi
o
n
er
s
with
v
alu
ab
le
in
s
ig
h
ts
to
m
ak
e
in
f
o
r
m
ed
d
ec
is
io
n
s
wh
en
s
elec
tin
g
an
L
L
M.
Ad
d
itio
n
ally
,
it
s
er
v
es a
s
a
r
ef
er
en
ce
f
o
r
f
i
n
e
-
t
u
n
in
g
s
u
m
m
ar
izatio
n
m
o
d
els to
b
etter
s
u
it sp
ec
if
ic
task
s
an
d
d
atasets
.
3.
E
XP
E
R
I
M
E
N
T
A
L
SE
T
UP
E
x
p
er
im
en
ts
wer
e
co
n
d
u
cted
f
o
r
ea
ch
L
L
M
u
s
in
g
a
f
ix
ed
te
m
p
er
atu
r
e
s
ettin
g
o
f
0
.
8
an
d
a
m
ax
im
u
m
to
k
en
len
g
t
h
o
f
8
0
.
T
h
e
s
tu
d
y
in
v
o
lv
ed
s
u
m
m
ar
izin
g
5
0
s
cien
tific
d
o
cu
m
e
n
ts
.
T
o
g
en
e
r
at
e
tex
t
s
u
m
m
ar
ies,
L
an
g
C
h
ain
an
d
Hu
g
g
in
g
Fa
ce
p
ip
elin
es
wer
e
u
tili
ze
d
f
o
r
p
r
o
m
p
t
en
g
in
ee
r
in
g
,
en
s
u
r
in
g
ac
cu
r
ac
y
an
d
ef
f
icien
cy
th
r
o
u
g
h
o
u
t
th
e
s
u
m
m
ar
izatio
n
p
r
o
ce
s
s
.
T
h
e
ex
p
er
im
en
ts
wer
e
ca
r
r
ied
o
u
t
u
tili
zin
g
a
Go
o
g
le
C
o
la
b
No
teb
o
o
k
,
wh
ich
was
eq
u
i
p
p
ed
with
T
4
GPUs
.
Ad
d
itio
n
ally
,
a
Kag
g
le
N
o
teb
o
o
k
with
a
GPU
P1
0
0
ac
ce
ler
ato
r
was
em
p
lo
y
ed
f
o
r
th
e
ex
p
er
im
e
n
ts
.
T
h
e
e
x
ec
u
tio
n
in
v
o
l
v
ed
t
h
e
u
tili
za
tio
n
o
f
a
n
Op
en
AI
API
k
e
y
an
d
th
e
Op
e
n
R
o
u
ter
p
la
y
g
r
o
u
n
d
f
o
r
r
ec
en
tly
lau
n
ch
e
d
m
o
d
e
ls
.
4.
I
NF
E
R
E
NC
E
WI
T
H
DIV
E
RSE
L
L
M
S
I
n
o
u
r
s
tu
d
y
o
n
ab
s
tr
ac
tiv
e
s
cien
tific
d
o
cu
m
en
t
s
u
m
m
ar
iz
atio
n
,
we
o
b
s
er
v
e
d
a
s
tr
o
n
g
co
r
r
elatio
n
b
etwe
en
r
ec
all
an
d
th
e
m
o
d
el's
ef
f
ec
tiv
en
ess
in
ca
p
tu
r
in
g
k
ey
p
o
in
ts
an
d
ess
en
tial
in
f
o
r
m
atio
n
.
R
ec
all,
also
k
n
o
wn
as
s
en
s
itiv
ity
,
ev
alu
ate
s
th
e
m
o
d
el'
s
ab
ilit
y
to
ac
c
u
r
a
tely
id
en
tify
an
d
in
co
r
p
o
r
ate
all
r
elev
an
t
d
etails
f
r
o
m
th
e
o
r
i
g
in
al
d
o
cu
m
en
t
in
to
th
e
g
en
er
ated
s
u
m
m
ar
y
.
A
h
ig
h
er
r
ec
all
in
d
icate
s
th
at
th
e
m
o
d
el
is
p
r
o
f
icien
t
in
ca
p
tu
r
in
g
im
p
o
r
tan
t
co
n
ten
t,
ev
en
if
it
m
ea
n
s
in
clu
d
in
g
s
o
m
e
n
o
n
-
ess
en
tial
d
etails.
G
iv
en
o
u
r
f
o
c
u
s
o
n
ex
tr
ac
tin
g
c
r
itical
co
n
ce
p
ts
a
n
d
in
s
ig
h
ts
f
r
o
m
s
cien
tific
d
o
cu
m
en
ts
,
p
r
io
r
itizin
g
r
ec
all
em
er
g
es
as
a
k
ey
o
b
jectiv
e,
en
s
u
r
i
n
g
co
m
p
r
eh
e
n
s
iv
e
co
v
er
a
g
e
o
f
r
elev
an
t in
f
o
r
m
atio
n
th
r
o
u
g
h
o
u
t th
e
s
u
m
m
ar
izatio
n
p
r
o
ce
s
s
.
Ho
wev
er
,
f
o
r
a
well
-
r
o
u
n
d
ed
ev
alu
atio
n
,
it
is
cr
u
cial
to
also
co
n
s
id
er
p
r
ec
is
io
n
an
d
th
e
F1
s
co
r
e.
Pre
cisi
o
n
ass
es
s
es
th
e
ac
cu
r
ac
y
o
f
th
e
g
en
er
ate
d
s
u
m
m
ar
y
b
y
d
eter
m
in
in
g
th
e
p
r
o
p
o
r
tio
n
o
f
co
r
r
ec
tly
id
en
tifie
d
r
elev
an
t
in
f
o
r
m
atio
n
r
elativ
e
to
th
e
to
tal
p
r
ed
icted
r
elev
an
t
co
n
ten
t.
T
h
e
F1
s
co
r
e,
ca
lcu
lated
as
th
e
h
ar
m
o
n
ic
m
ea
n
o
f
p
r
ec
is
io
n
a
n
d
r
ec
all,
o
f
f
er
s
a
b
alan
ce
d
m
ea
s
u
r
e
b
y
ac
co
u
n
tin
g
f
o
r
b
o
th
f
alse
p
o
s
itiv
es
an
d
f
alse
n
eg
ativ
es.
B
y
in
co
r
p
o
r
ati
n
g
r
ec
all,
p
r
ec
is
io
n
,
an
d
th
e
F
1
s
co
r
e,
th
is
co
m
p
r
eh
en
s
iv
e
a
p
p
r
o
ac
h
alig
n
s
with
o
u
r
o
b
jectiv
e
o
f
ex
tr
ac
tin
g
k
e
y
co
n
ce
p
ts
an
d
in
s
ig
h
ts
f
r
o
m
s
cien
tific
d
o
cu
m
en
ts
wh
ile
en
s
u
r
in
g
th
e
ac
cu
r
ac
y
an
d
r
elev
a
n
ce
o
f
th
e
s
u
m
m
ar
ie
s
.
W
e
f
o
u
n
d
th
at
m
o
d
els
s
u
ch
as
'
C
lau
d
e
v
2
.
1
,
'
'
P
PLX
7
0
B
On
lin
e,
'
an
d
'
Mistra
l
7
B
I
n
s
tr
u
ct
'
d
em
o
n
s
tr
ate
ex
ce
p
tio
n
ally
h
ig
h
r
ec
all
v
alu
es,
in
d
icatin
g
th
eir
p
r
o
f
icien
cy
in
e
f
f
ec
ti
v
ely
ex
tr
ac
tin
g
a
s
u
b
s
tan
tial
am
o
u
n
t
o
f
r
elev
a
n
t
in
f
o
r
m
atio
n
f
r
o
m
t
h
e
s
o
u
r
ce
t
ex
t.
T
h
is
p
r
o
f
icien
cy
is
s
u
p
p
o
r
ted
b
y
r
em
ar
k
a
b
le
wo
r
d
o
v
er
lap
ac
c
u
r
ac
y
,
as
s
h
o
wn
in
Fig
u
r
e
1
,
h
ig
h
lig
h
tin
g
th
eir
p
r
ec
is
io
n
i
n
s
u
m
m
ar
izatio
n
th
r
o
u
g
h
s
u
b
s
tan
tial w
o
r
d
alig
n
m
en
t.
T
h
ese
m
o
d
els
ar
e
ef
f
ec
tiv
e
at
m
in
im
izin
g
th
e
o
m
is
s
io
n
o
f
im
p
o
r
tan
t
d
etails
in
th
eir
s
u
m
m
ar
ies.
C
o
n
v
er
s
ely
,
m
o
d
els
with
m
o
d
er
ate
r
ec
all,
lik
e
'
g
p
t2
-
x
l,'
'
f
alco
n
-
7b
-
in
s
tr
u
ct,
'
an
d
'
C
l
au
d
e
v
2
.
0
,
'
ca
p
tu
r
e
r
elev
an
t
in
f
o
r
m
atio
n
r
ea
s
o
n
a
b
ly
well
b
u
t
d
o
n
o
t
ac
h
iev
e
th
e
h
ig
h
est
r
ec
all
v
alu
es.
T
h
ese
m
o
d
els
b
alan
ce
p
r
ec
is
io
n
an
d
r
ec
all
in
th
e
s
u
m
m
ar
izatio
n
p
r
o
ce
s
s
.
Mo
d
els
with
lo
w
r
ec
all,
s
u
ch
as
'
t5
-
s
m
all,
'
'
p
eg
asu
s
-
x
s
u
m
,
'
an
d
'
m
p
t
-
7b
-
i
n
s
tr
u
ct,
'
m
ay
h
av
e
lim
itatio
n
s
in
ca
p
tu
r
in
g
a
s
u
b
s
tan
tial
p
o
r
tio
n
o
f
r
el
ev
an
t
in
f
o
r
m
atio
n
,
p
o
ten
tially
m
is
s
in
g
im
p
o
r
tan
t
d
etails
in
th
eir
s
u
m
m
ar
ies.
Fi
g
u
r
es
2
to
4
illu
s
tr
ate
th
e
r
ec
a
ll
an
d
F1
s
co
r
es
o
f
th
ese
m
o
d
els f
o
r
R
OUGE
-
1
,
R
OUGE
-
2
,
an
d
R
OUGE
-
L.
T
h
e
ev
alu
atio
n
m
etr
ics
p
r
esen
ted
in
T
ab
le
1
r
ev
ea
l
th
at
'C
lau
d
e
v
2
.
1
,
'
'PPLX
7
0
B
O
n
lin
e,
'
an
d
'
Mi
s
tr
al
7
B
I
n
s
tr
u
ct
'
ex
h
ib
it
ex
ce
p
tio
n
al
B
L
E
U
s
co
r
es,
r
ef
l
ec
tin
g
th
eir
p
r
o
f
icien
cy
in
g
e
n
er
atin
g
s
u
m
m
ar
ies
th
at
clo
s
ely
m
atch
r
ef
er
e
n
ce
tex
ts
ac
r
o
s
s
v
ar
io
u
s
n
-
g
r
am
o
r
d
er
s
.
Ad
d
itio
n
ally
,
B
E
R
T
s
co
r
es
h
ig
h
lig
h
t
t
h
e
b
alan
ce
b
etwe
en
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
,
with
s
o
m
e
m
o
d
els
s
h
o
win
g
e
x
ce
p
tio
n
al
o
v
er
all
p
er
f
o
r
m
a
n
ce
.
No
tab
ly
,
'
C
lau
d
e
v
1
'
p
r
esen
ts
a
co
m
p
ar
ativ
ely
lo
wer
B
E
R
T
s
co
r
e,
s
u
g
g
esti
n
g
a
p
o
ten
ti
al
tr
ad
e
-
o
f
f
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all.
Ov
e
r
all,
th
ese
m
etr
ics
p
r
o
v
id
e
a
n
u
a
n
ce
d
u
n
d
er
s
tan
d
in
g
o
f
ea
ch
L
L
M'
s
s
tr
en
g
th
s
an
d
wea
k
n
ess
es in
tex
t su
m
m
ar
izatio
n
task
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
I
n
ve
s
tig
a
tin
g
th
e
r
ec
a
ll e
fficie
n
cy
in
a
b
s
tr
a
ctive
s
u
mma
r
iz
a
t
io
n
:
a
n
ex
p
erimen
ta
l
…
(
S
u
r
a
b
h
i A
n
u
r
a
d
h
a
)
451
Fig
u
r
e
1
.
W
o
r
d
o
v
e
r
lap
ac
cu
r
a
cy
ac
r
o
s
s
d
if
f
er
e
n
t L
L
Ms
Fig
u
r
e
2
.
R
OUGE
-
1
r
ec
all
an
d
F1
s
co
r
e
v
alu
es a
m
o
n
g
d
if
f
e
r
en
t L
L
Ms
Fig
u
r
e
3
.
R
OUGE
-
2
r
ec
all
an
d
F1
s
co
r
e
v
alu
es a
m
o
n
g
d
if
f
e
r
en
t L
L
Ms
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
446
-
4
5
4
452
Fig
u
r
e
4
.
R
OUGE
-
L
r
ec
all
an
d
F1
s
co
r
e
v
alu
es a
m
o
n
g
d
if
f
e
r
en
t L
L
Ms
T
ab
le
1
.
B
L
E
U
an
d
B
E
R
T
s
co
r
es
co
m
p
ar
ed
ac
r
o
s
s
d
if
f
er
e
n
t
L
L
Ms
LLM
mo
d
e
l
B
LEU
-
1
B
LEU
-
2
B
LEU
-
3
B
LEU
-
4
B
LEU
B
ER
T
s
c
o
r
e
(
*
P
/
R
/
F
1
)
B
a
r
t
-
l
a
r
g
e
-
c
n
n
0
.
3
7
0
.
1
2
3
0
.
0
4
1
0
.
0
1
3
8
.
8
4
E
-
2
3
2
0
.
8
9
5
/
0
.
9
0
1
/
0
.
8
9
8
T5
-
sma
l
l
0
.
4
0
.
1
3
3
0
.
0
4
4
0
.
0
1
4
1
.
0
2
E
-
2
3
1
0
.
8
6
0
/
0
.
8
5
7
/
0
.
8
5
9
D
i
st
i
l
b
a
r
t
-
c
n
n
-
12
-
6
0
.
4
0
.
1
3
3
0
.
0
4
4
0
.
0
1
4
1
.
0
2
E
-
2
3
1
0
.
8
9
2
/
0
.
8
8
5
/
0
.
8
8
9
P
e
g
a
s
u
s
-
x
su
m
0
.
3
7
2
0
.
1
2
4
0
.
0
4
1
0
.
0
1
3
8
.
9
7
E
-
2
3
2
0
.
8
6
5
/
0
.
8
8
1
/
0
.
8
7
3
G
p
t
2
-
xl
0
.
3
5
5
0
.
1
1
8
0
.
0
3
9
0
.
0
1
3
7
.
7
8
E
-
2
3
2
0
.
9
0
1
/
0
.
8
7
5
/
0
.
8
8
8
Te
x
t
-
d
a
v
i
n
c
i
-
0
0
3
0
.
3
5
2
0
.
1
1
7
0
.
0
3
9
0
.
0
1
3
7
.
5
4
E
-
2
3
2
0
.
9
3
1
/
0
.
9
0
1
/
0
.
9
1
6
G
p
t
-
3
.
5
-
t
u
r
b
o
0
.
3
5
1
0
.
1
1
6
0
.
0
3
7
0
.
0
1
2
6
.
8
5
E
-
2
3
2
0
.
9
2
2
/
0
.
8
9
7
/
0
.
9
0
9
G
p
t
-
4
0
.
3
5
5
0
.
1
1
8
0
.
3
9
5
0
.
0
1
3
7
.
7
8
E
-
2
3
2
0
.
9
2
5
/
0
.
8
9
8
/
0
.
9
1
1
M
p
t
-
7b
-
i
n
st
r
u
c
t
0
.
3
4
6
0
.
1
1
5
0
.
0
3
8
0
.
0
1
2
6
.
8
5
E
-
2
3
2
0
.
8
9
0
/
0
.
8
4
/
0
.
8
6
5
F
a
l
c
o
n
-
7b
-
i
n
st
r
u
c
t
0
.
3
5
3
0
.
1
1
7
0
.
0
3
9
0
.
0
1
3
7
.
6
0
E
-
2
3
2
0
.
9
3
3
/
0
.
9
0
7
/
0
.
9
2
0
F
l
a
n
-
t5
-
x
x
l
0
.
3
5
5
0
.
1
1
8
0
.
0
3
9
0
.
0
1
3
7
.
7
8
E
-
2
3
2
0
.
9
0
1
/
0
.
8
7
7
/
0
.
8
8
9
M
i
s
t
r
a
l
7
B
I
n
st
r
u
c
t
0
.
3
4
7
0
.
1
1
5
0
.
0
3
8
0
.
0
1
2
6
.
9
2
E
-
2
3
2
0
.
9
4
9
/
0
.
8
8
8
/
0
.
9
1
7
M
y
t
h
o
M
i
s
t
7
B
0
.
3
5
2
0
.
1
1
6
0
.
0
3
7
0
.
0
1
3
7
.
8
3
E
-
2
2
9
0
.
9
1
1
/
0
.
8
9
2
/
0
.
9
0
1
M
y
t
h
o
M
a
x
1
3
B
8
k
0
.
3
5
6
0
.
1
1
5
0
.
0
3
7
0
.
0
1
2
7
.
8
8
E
-
2
3
2
0
.
9
0
9
/
0
.
8
9
9
/
0
.
9
0
4
Ll
a
m
a
v
2
1
3
B
C
h
a
t
0
.
3
4
2
0
.
1
1
0
.
0
3
8
0
.
0
1
2
6
.
2
3
E
-
2
3
2
0
.
9
1
6
/
0
.
8
3
3
/
0
.
8
7
3
C
o
d
e
Ll
a
ma
3
4
B
v
2
0
.
3
5
2
0
.
1
1
7
0
.
0
3
9
0
.
0
1
3
7
.
4
9
E
-
2
3
2
0
.
9
2
/
0
.
8
9
8
/
0
.
9
1
2
P
a
LM
2
C
h
a
t
0
.
3
5
8
0
.
1
1
9
0
.
0
3
9
0
.
0
1
3
7
.
9
9
E
-
2
3
2
0
.
8
9
8
/
0
.
8
8
2
/
0
.
8
9
0
P
a
LM
2
C
h
a
t
3
2
k
0
.
3
5
4
0
.
1
1
8
0
.
0
3
9
0
.
0
1
3
7
.
7
2
E
-
2
3
2
0
.
9
3
2
/
0
.
9
0
1
/
0
.
9
1
6
C
l
a
u
d
e
v
1
0
.
3
5
4
0
.
1
1
7
0
.
0
3
9
0
.
0
1
3
7
.
7
4
E
-
2
3
2
0
.
9
0
8
/
0
.
5
7
7
/
0
.
8
9
4
C
l
a
u
d
e
v
2
.
0
0
.
3
5
1
0
.
1
1
7
0
.
0
3
9
0
.
0
1
3
7
.
3
6
E
-
2
3
2
0
.
9
3
8
/
0
.
8
9
3
/
0
.
9
1
5
C
l
a
u
d
e
v
2
.
1
0
.
3
5
0
.
1
1
6
0
.
0
3
8
0
.
0
1
2
7
.
2
0
E
-
2
3
2
0
.
9
6
8
/
0
.
9
0
2
/
0
.
9
3
4
G
o
l
i
a
t
h
1
2
0
B
0
.
3
5
4
0
.
1
1
8
0
.
0
3
9
0
.
0
1
3
7
.
6
5
E
-
2
3
2
0
.
9
5
7
/
0
.
9
2
0
/
0
.
9
3
8
H
e
r
mes
7
0
B
0
.
3
5
8
0
.
1
1
9
0
.
0
3
9
0
.
0
1
3
7
.
8
9
E
-
2
3
2
0
.
9
1
1
/
0
.
8
8
7
/
0
.
8
9
9
P
P
LX
7
0
B
O
n
l
i
n
e
0
.
3
4
8
0
.
1
1
6
0
.
0
3
8
0
.
0
1
2
7
.
0
7
E
-
2
3
2
0
.
9
5
5
/
0
.
8
9
4
/
0
.
9
2
5
To
p
p
y
M
7
B
0
.
3
4
6
0
.
1
1
7
0
.
0
3
7
0
.
0
1
3
7
.
3
7
E
-
2
3
2
0
.
9
1
6
/
0
.
9
0
1
/
0
.
9
0
9
*P
-
P
r
e
c
i
si
o
n
,
R
-
R
e
c
a
l
l
,
F
1
-
F
1
s
c
o
r
e
5.
CO
NCLU
SI
O
N
AND
F
U
T
U
RE
WO
RK
T
h
is
s
tu
d
y
in
v
esti
g
ated
th
e
u
s
e
o
f
v
ar
io
u
s
LLMs
f
o
r
s
u
m
m
ar
izin
g
s
cien
tific
d
o
cu
m
e
n
ts
.
W
h
ile
p
r
ev
io
u
s
r
esear
c
h
h
as
ex
p
l
o
r
ed
d
if
f
er
e
n
t
s
u
m
m
ar
izatio
n
te
ch
n
iq
u
es,
it
h
as
n
o
t
s
p
ec
if
ica
lly
ad
d
r
ess
ed
th
eir
im
p
ac
t
o
n
s
u
m
m
ar
y
co
h
e
r
en
ce
an
d
th
e
th
o
r
o
u
g
h
in
cl
u
s
io
n
o
f
k
ey
d
etails.
Ou
r
f
i
n
d
in
g
s
r
ev
ea
l
a
s
tr
o
n
g
co
r
r
elatio
n
b
etwe
en
r
ec
all
an
d
th
e
q
u
ality
o
f
th
e
g
en
er
ate
d
s
u
m
m
ar
ies.
Ad
v
an
c
ed
L
L
Ms
s
u
ch
as
'C
lau
d
e
v
2
.
1
,
'
'
P
PLX
7
0
B
On
lin
e,
'
an
d
'
M
is
tr
al
7
B
I
n
s
tr
u
ct'
d
em
o
n
s
tr
ated
g
r
ea
ter
ef
f
ec
tiv
en
ess
th
a
n
ea
r
lier
m
o
d
els
in
ca
p
tu
r
in
g
ess
en
tial
in
f
o
r
m
ati
o
n
.
Ad
d
itio
n
ally
,
r
ec
en
t
o
b
s
er
v
atio
n
s
in
d
icate
th
at
im
p
r
o
v
ed
s
u
m
m
ar
izatio
n
q
u
ality
is
ass
o
ciate
d
w
ith
en
h
an
ce
d
m
o
d
el
r
ec
all,
r
ath
er
th
a
n
m
er
ely
an
in
cr
ea
s
e
in
th
e
n
u
m
b
er
o
f
g
e
n
er
ate
d
to
k
en
s
.
Ou
r
r
esu
lts
p
r
o
v
id
e
co
m
p
ellin
g
e
v
id
en
ce
t
h
at
h
ig
h
er
r
ec
all
m
etr
ics
co
n
tr
ib
u
te
to
m
o
r
e
ef
f
ec
tiv
e
s
u
m
m
ar
ies.
Fu
tu
r
e
r
esear
ch
co
u
ld
ex
p
lo
r
e
en
s
em
b
le
ap
p
r
o
a
ch
es
to
h
ar
n
ess
th
e
s
tr
en
g
th
s
o
f
m
u
ltip
le
m
o
d
els
f
o
r
ev
e
n
b
etter
s
u
m
m
ar
izatio
n
o
u
tco
m
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
I
n
ve
s
tig
a
tin
g
th
e
r
ec
a
ll e
fficie
n
cy
in
a
b
s
tr
a
ctive
s
u
mma
r
iz
a
t
io
n
:
a
n
ex
p
erimen
ta
l
…
(
S
u
r
a
b
h
i A
n
u
r
a
d
h
a
)
453
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Su
r
ab
h
i A
n
u
r
ad
h
a
✓
✓
✓
✓
✓
✓
✓
✓
Ma
r
th
a
Sh
esh
ik
ala
✓
✓
✓
✓
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
T
h
e
d
ata
t
h
at
s
u
p
p
o
r
t
th
e
f
i
n
d
in
g
s
o
f
th
is
s
tu
d
y
ar
e
av
a
ilab
le
f
r
o
m
th
e
c
o
r
r
esp
o
n
d
in
g
au
th
o
r
,
[
Su
r
ab
h
i]
,
u
p
o
n
r
ea
s
o
n
ab
le
r
e
q
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
G
h
a
d
i
mi
a
n
d
H
.
B
e
i
g
y
,
“
H
y
b
r
i
d
mu
l
t
i
-
d
o
c
u
me
n
t
su
mm
a
r
i
z
a
t
i
o
n
u
s
i
n
g
p
r
e
-
t
r
a
i
n
e
d
l
a
n
g
u
a
g
e
mo
d
e
l
s,
”
E
x
p
e
r
t
S
y
st
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s,
v
o
l
.
1
9
2
,
A
p
r
.
2
0
2
2
,
p
p
.
1
1
6
2
9
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
1
.
1
1
6
2
9
2
.
[
2
]
L.
B
a
sy
a
l
a
n
d
M
.
S
a
n
g
h
v
i
,
“
T
e
x
t
s
u
mm
a
r
i
z
a
t
i
o
n
u
si
n
g
l
a
r
g
e
l
a
n
g
u
a
g
e
m
o
d
e
l
s
:
a
c
o
mp
a
r
a
t
i
v
e
s
t
u
d
y
o
f
M
P
T
-
7b
-
i
n
st
r
u
c
t
,
F
a
l
c
o
n
-
7b
-
i
n
s
t
r
u
c
t
,
a
n
d
O
p
e
n
A
I
C
h
a
t
-
G
P
T
M
o
d
e
l
s
,
”
a
rX
i
v
p
r
e
p
r
i
n
t
,
O
c
t
.
2
0
2
3
,
a
r
X
i
v
:
2
3
1
0
.
1
0
4
4
9
.
[
3
]
M
.
Le
w
i
s,
e
t
a
l
.
,
“
B
A
R
T
:
d
e
n
o
i
s
i
n
g
se
q
u
e
n
c
e
-
to
-
s
e
q
u
e
n
c
e
p
r
e
-
t
r
a
i
n
i
n
g
f
o
r
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
g
e
n
e
r
a
t
i
o
n
,
t
r
a
n
sl
a
t
i
o
n
,
a
n
d
c
o
m
p
r
e
h
e
n
si
o
n
,
”
a
r
Xi
v
p
re
p
ri
n
t
,
O
c
t
.
2
0
1
9
,
a
r
X
i
v
:
1
9
1
0
.
1
3
4
6
1
.
[
4
]
O
p
e
n
A
I
,
G
P
T
-
3
.
5
,
t
e
x
t
-
d
a
v
i
n
c
i
-
0
0
3
.
h
t
t
p
s:
/
/
p
l
a
t
f
o
r
m.
o
p
e
n
a
i
.
c
o
m/
d
o
c
s/
mo
d
e
l
s/
g
p
t
-
3
-
5
.
(
A
c
c
e
ss
e
d
:
D
e
c
,
4
2
0
2
3
).
[
5
]
K
.
W
i
g
g
e
r
s
,
“
O
p
e
n
A
I
r
e
l
e
a
s
e
s
G
P
T
-
4
,
a
m
u
l
t
i
mo
d
a
l
A
I
t
h
a
t
i
t
c
l
a
i
m
s
i
s
s
t
a
t
e
-
of
-
t
h
e
-
a
r
t
,”
T
e
c
h
C
r
u
n
c
h
,
2
0
2
3
,
A
c
c
e
ss
e
d
1
5
M
a
r
c
h
2
0
2
3
.
[
6
]
M
o
s
a
i
c
M
L
N
LP
T
e
a
m
(
2
0
2
3
)
,
“
I
n
t
r
o
d
u
c
i
n
g
M
P
T
-
7
B
:
a
n
e
w
s
t
a
n
d
a
r
d
f
o
r
o
p
e
n
-
so
u
r
c
e
,
c
o
mm
e
r
c
i
a
l
l
y
u
sa
b
l
e
L
LM
s
,
”
h
t
t
p
s
:
/
/
w
w
w
.
mo
sai
c
ml
.
c
o
m/
b
l
o
g
/
mp
t
-
7
b
.
A
c
c
e
ss
e
d
4
D
e
c
e
mb
e
r
2
0
2
3
[
7
]
G
.
P
e
n
e
d
o
e
t
a
l
.
,
“
T
h
e
R
e
f
i
n
e
d
W
e
b
d
a
t
a
se
t
f
o
r
F
a
l
c
o
n
LL
M
:
o
u
t
p
e
r
f
o
r
m
i
n
g
c
u
r
a
t
e
d
c
o
r
p
o
r
a
w
i
t
h
w
e
b
d
a
t
a
o
n
l
y
,
”
A
d
v
a
n
c
e
s
i
n
N
e
u
ra
l
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
S
y
s
t
e
m
s,
v
o
l
.
3
6
,
D
e
c
.
2
0
2
3
,
p
p
.
7
9
1
5
5
-
7
9
1
7
2
.
[
8
]
A
.
Q
.
Ji
a
n
g
,
A
.
S
a
b
l
a
y
r
o
l
l
e
s
,
A
.
M
e
n
sch
,
C
.
B
a
mf
o
r
d
,
D
.
S
.
C
h
a
p
l
o
t
,
a
n
d
W
.
E
.
S
a
y
e
d
,
“
M
i
s
t
r
a
l
7
B
,
”
a
rX
i
v
p
re
p
ri
n
t
,
O
c
t
.
2
0
2
3
,
a
r
X
i
v
:
2
3
1
0
.
0
6
8
2
5
.
[
9
]
H
.
To
u
v
r
o
n
e
t
a
l
.
,
“
L
l
a
m
a
2
:
o
p
e
n
f
o
u
n
d
a
t
i
o
n
a
n
d
f
i
n
e
-
t
u
n
e
d
c
h
a
t
m
o
d
e
l
s
,
”
a
rXi
v
p
re
p
r
i
n
t
,
J
u
l
.
2
0
2
3
,
a
r
X
i
v
:
2
3
0
7
.
0
9
2
8
8
.
[
1
0
]
O
.
K
h
a
t
t
a
b
e
t
a
l
.
,
"
D
S
P
y
:
c
o
m
p
i
l
i
n
g
d
e
c
l
a
r
a
t
i
v
e
l
a
n
g
u
a
g
e
m
o
d
e
l
c
a
l
l
s
i
n
t
o
sel
f
-
i
mp
r
o
v
i
n
g
p
i
p
e
l
i
n
e
s,"
a
r
Xi
v
p
re
p
ri
n
t
,
2
0
2
3
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
2
3
1
0
.
0
3
7
1
4
.
[
1
1
]
T.
S
.
W
a
n
g
a
n
d
A
.
S
.
G
o
r
d
o
n
,
"
P
l
a
y
i
n
g
st
o
r
y
c
r
e
a
t
i
o
n
g
a
mes
w
i
t
h
l
a
r
g
e
l
a
n
g
u
a
g
e
m
o
d
e
l
s:
e
x
p
e
r
i
me
n
t
s
w
i
t
h
G
P
T
-
3
.
5
,
"
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
t
e
r
a
c
t
i
v
e
D
i
g
i
t
a
l
S
t
o
ry
t
e
l
l
i
n
g
,
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
,
p
p
.
2
9
7
-
3
0
5
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
9
7
8
-
3
-
0
3
1
-
4
7
6
5
8
-
7
_
2
8
.
[
1
2
]
J.
C
u
s
i
d
ó
,
L.
S
o
l
é
-
V
i
l
a
r
ó
,
P
.
M
a
r
t
i
-
P
u
i
g
,
a
n
d
J.
S
o
l
é
-
C
a
sa
l
s,
“
A
ssess
i
n
g
t
h
e
c
a
p
a
b
i
l
i
t
y
o
f
a
d
v
a
n
c
e
d
A
I
mo
d
e
l
s
i
n
c
a
r
d
i
o
v
a
sc
u
l
a
r
sy
m
p
t
o
m re
c
o
g
n
i
t
i
o
n
:
a
c
o
m
p
a
r
a
t
i
v
e
s
t
u
d
y
,
”
Ap
p
l
i
e
d
S
c
i
e
n
c
e
s
,
v
o
l
.
1
4
,
n
o
.
1
8
,
p
.
8
4
4
0
,
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
a
p
p
1
4
1
8
8
4
4
0
.
[
1
3
]
G
r
e
p
h
y
,
M
y
t
h
o
M
i
st
7
B
,
g
r
e
p
h
y
/
m
y
t
h
o
m
y
s
t
-
7
b
(
2
0
2
3
)
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s:
/
/
o
p
e
n
r
o
u
t
e
r
.
a
i
/
m
o
d
e
l
s
/
g
r
y
p
h
e
/
m
y
t
h
o
mi
s
t
-
7b
.
(
A
c
c
e
sse
d
:
D
e
c
,
4
2
0
2
3
)
[
1
4
]
M
.
R
.
C
.
Q
a
z
a
n
i
,
H
.
A
sa
d
i
,
S
.
M
o
h
a
m
e
d
,
S
.
N
a
h
a
v
a
n
d
i
,
J.
W
i
n
t
e
r
a
n
d
K
.
R
o
sar
i
o
,
"
A
r
e
a
l
-
t
i
m
e
m
o
t
i
o
n
c
o
n
t
r
o
l
t
r
a
c
k
i
n
g
mec
h
a
n
i
sm
f
o
r
sat
e
l
l
i
t
e
t
r
a
c
k
i
n
g
a
n
t
e
n
n
a
u
s
i
n
g
ser
i
a
l
r
o
b
o
t
,
"
2
0
2
1
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
S
y
st
e
m
s,
Ma
n
,
a
n
d
C
y
b
e
r
n
e
t
i
c
s (S
MC
)
,
M
e
l
b
o
u
r
n
e
,
A
u
s
t
r
a
l
i
a
,
2
0
2
1
,
p
p
.
1
0
4
9
-
1
0
5
5
,
d
o
i
:
1
0
.
1
1
0
9
/
S
M
C
5
2
4
2
3
.
2
0
2
1
.
9
6
5
8
9
0
9
.
[
1
5
]
G
o
o
g
l
e
Te
a
m
,
P
a
L
M
2
C
o
d
e
C
h
a
t
3
2
k
,
g
o
o
g
l
e
/
p
a
l
m
-
2
-
c
o
d
e
c
h
a
t
-
b
i
so
n
-
3
2
k
(
2
0
2
3
)
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
o
p
e
n
r
o
u
t
e
r
.
a
i
/
mo
d
e
l
s
/
g
o
o
g
l
e
/
p
a
l
m
-
2
-
c
o
d
e
c
h
a
t
-
b
i
s
o
n
-
3
2
k
.
(
A
c
c
e
ss
e
d
:
D
e
c
,
4
2
0
2
3
)
[
1
6
]
L.
D
e
n
g
e
t
a
l
.
,
“
E
v
a
l
u
a
t
i
o
n
o
f
l
a
r
g
e
l
a
n
g
u
a
g
e
mo
d
e
l
s
i
n
b
r
e
a
s
t
c
a
n
c
e
r
c
l
i
n
i
c
a
l
sce
n
a
r
i
o
s:
a
c
o
m
p
a
r
a
t
i
v
e
a
n
a
l
y
s
i
s
b
a
s
e
d
o
n
C
h
a
t
G
P
T
-
3
.
5
,
C
h
a
t
G
P
T
-
4
.
0
,
a
n
d
C
l
a
u
d
e
2
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
S
u
r
g
e
ry
,
v
o
l
.
1
1
0
,
n
o
.
4
,
A
p
r
.
2
0
2
4
,
p
p
.
1
9
4
1
-
1
9
5
0
,
d
o
i
:
1
0
.
1
0
9
7
/
JS
9
.
0
0
0
0
0
0
0
0
0
0
0
0
1
0
6
6
.
[
1
7
]
A
n
t
h
r
o
p
i
c
Te
a
m,
A
n
t
h
r
o
p
i
c
C
o
m
p
l
e
t
i
o
n
M
o
d
e
l
s
(
2
0
2
3
)
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
c
l
a
r
i
f
a
i
.
c
o
m/
a
n
t
h
r
o
p
i
c
/
c
o
mp
l
e
t
i
o
n
/
mo
d
e
l
s
.
A
c
c
e
ss
e
d
4
D
e
c
e
mb
e
r
2
0
2
3
[
1
8
]
Y
.
O
u
,
Z.
H
u
i
,
T.
Z
h
o
u
,
Y
.
C
a
i
,
a
n
d
J
.
Li
,
“
Ll
a
ma
2
-
1
3
b
-
b
a
se
d
N
EF
T
f
i
n
e
-
t
u
n
i
n
g
f
o
r
f
i
n
a
n
c
i
a
l
se
n
t
i
me
n
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
2
4
G
u
a
n
g
d
o
n
g
-
H
o
n
g
K
o
n
g
-
M
a
c
a
o
G
r
e
a
t
e
r
B
a
y
Ar
e
a
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
D
i
g
i
t
a
l
E
c
o
n
o
m
y
a
n
d
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
Ja
n
.
2
0
2
4
,
p
p
.
6
4
1
-
6
4
4
,
d
o
i
:
1
0
.
1
1
4
5
/
3
6
7
5
4
1
7
.
3
6
7
5
5
2
3
.
[
1
9
]
W
.
L.
C
h
i
a
n
g
e
t
a
l
.
,
“
C
h
a
t
b
o
t
a
r
e
n
a
:
a
n
o
p
e
n
p
l
a
t
f
o
r
m
f
o
r
e
v
a
l
u
a
t
i
n
g
L
LM
s
b
y
h
u
m
a
n
p
r
e
f
e
r
e
n
c
e
,
”
Fo
r
t
y
-
f
i
rst
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
,
2
0
2
4
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
2
4
0
3
.
0
4
1
3
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
39
,
No
.
1
,
J
u
ly
20
25
:
446
-
4
5
4
454
[
2
0
]
M
.
S
h
a
f
i
q
u
e
,
G
.
M
u
mt
a
z
,
S
.
Z.
A
h
m
a
d
,
a
n
d
S
.
I
q
b
a
l
,
“
A
c
o
mp
a
r
a
t
i
v
e
a
n
a
l
y
si
s
o
f
A
I
c
h
a
t
b
o
t
p
e
r
f
o
r
ma
n
c
e
i
n
I
o
T
e
n
v
i
r
o
n
m
e
n
t
s,
”
J
o
u
rn
a
l
o
f
C
o
m
p
u
t
i
n
g
& B
i
o
m
e
d
i
c
a
l
I
n
f
o
rm
a
t
i
c
s
,
v
o
l
.
7
,
n
o
.
2
,
2
0
2
4
.
[
2
1
]
I
.
C
a
c
h
o
l
a
,
K
.
L
o
,
A
.
C
o
h
a
n
,
a
n
d
D
.
S
.
W
e
l
d
,
“
TLD
R
:
e
x
t
r
e
me
s
u
mm
a
r
i
z
a
t
i
o
n
o
f
sci
e
n
t
i
f
i
c
d
o
c
u
m
e
n
t
s,
”
a
r
Xi
v
p
re
p
r
i
n
t
,
A
p
r
.
2
0
2
0
,
a
r
X
i
v
:
2
0
0
4
.
1
5
0
1
1
.
[
2
2
]
K
.
P
a
p
i
n
e
n
i
,
S
.
R
o
u
k
o
s,
T
.
W
a
r
d
,
a
n
d
W
.
J.
Zh
u
,
“
B
LEU
:
a
me
t
h
o
d
f
o
r
a
u
t
o
mat
i
c
e
v
a
l
u
a
t
i
o
n
o
f
m
a
c
h
i
n
e
t
r
a
n
s
l
a
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
4
0
t
h
A
n
n
u
a
l
M
e
e
t
i
n
g
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s,
Ju
l
.
2
0
0
2
,
p
p
.
3
1
1
-
3
1
8
.
[
2
3
]
A
.
R
.
F
a
b
b
r
i
,
W
.
K
r
y
ści
ń
s
k
i
,
B
.
M
c
C
a
n
n
,
C
.
X
i
o
n
g
,
R
.
S
o
c
h
e
r
,
a
n
d
D
.
R
a
d
e
v
,
“
S
u
mm
e
v
a
l
:
r
e
-
e
v
a
l
u
a
t
i
n
g
s
u
mm
a
r
i
z
a
t
i
o
n
e
v
a
l
u
a
t
i
o
n
,
”
T
r
a
n
sa
c
t
i
o
n
s
o
f
t
h
e
Ass
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
st
i
c
s
,
v
o
l
.
9
,
A
p
r
.
2
0
2
1
,
p
p
.
3
9
1
-
4
0
9
.
[
2
4
]
C
.
Y
.
Li
n
,
“
R
o
u
g
e
:
a
p
a
c
k
a
g
e
f
o
r
a
u
t
o
mat
i
c
e
v
a
l
u
a
t
i
o
n
o
f
su
mm
a
r
i
e
s,
”
i
n
T
e
x
t
s
u
m
m
a
r
i
z
a
t
i
o
n
b
r
a
n
c
h
e
s
o
u
t
,
Ju
l
.
2
0
0
4
,
p
p
.
7
4
-
8
1
.
[
2
5
]
T.
Z
h
a
n
g
,
V
.
K
i
s
h
o
r
e
,
F
.
W
u
,
K
.
Q
.
W
e
i
n
b
e
r
g
e
r
,
a
n
d
Y
.
A
r
t
z
i
,
“
B
ER
T
sco
r
e
:
e
v
a
l
u
a
t
i
n
g
t
e
x
t
g
e
n
e
r
a
t
i
o
n
w
i
t
h
B
E
R
T
,”
a
rX
i
v
p
re
p
ri
n
t
,
A
p
r
.
2
0
1
9
,
a
r
X
i
v
:
1
9
0
4
.
0
9
6
7
5
.
[
2
6
]
W
.
Y
u
a
n
,
G
.
N
e
u
b
i
g
,
a
n
d
P
.
Li
u
,
“
B
A
R
T
sc
o
r
e
:
e
v
a
l
u
a
t
i
n
g
g
e
n
e
r
a
t
e
d
t
e
x
t
a
s
t
e
x
t
g
e
n
e
r
a
t
i
o
n
,
”
i
n
Ad
v
.
N
e
u
ra
l
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ssi
n
g
S
y
s
t
e
m
s
C
o
n
f
e
r
e
n
c
e
(
N
e
u
r
I
PS
)
,
v
o
l
.
3
4
,
A
p
r
.
2
0
2
1
,
p
p
.
2
7
2
6
3
-
2
7
2
7
7
,
d
o
i
:
1
0
.
4
8
5
5
0
/
a
r
X
i
v
.
1
9
0
4
.
0
9
6
7
5
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
S
u
r
a
b
h
i
Anu
r
a
d
h
a
is
a
re
se
a
rc
h
sc
h
o
lar
a
t
t
h
e
S
c
h
o
o
l
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
Artifi
c
ial
In
telli
g
e
n
c
e
,
S
R
Un
i
v
e
r
sity
,
Wara
n
g
a
l.
S
h
e
a
lso
se
rv
e
s
a
s
a
n
As
so
c
iate
P
ro
fe
ss
o
r
in
D
e
p
a
rtme
n
t
o
f
CS
E
(AIM
L)
a
t
Ke
sh
a
v
M
e
m
o
rial
I
n
stit
u
te
o
f
Tec
h
n
o
l
o
g
y
,
H
y
d
e
ra
b
a
d
,
In
d
ia
.
Wi
th
o
v
e
r
two
d
e
c
a
d
e
s
o
f
e
x
p
e
rien
c
e
in
e
d
u
c
a
ti
o
n
a
n
d
a
d
m
in
is
tratio
n
,
sh
e
sp
e
c
ialize
s
in
a
rti
ficia
l
in
telli
g
e
n
c
e
,
m
a
c
h
in
e
l
e
a
rn
in
g
,
d
e
e
p
lea
rn
i
n
g
,
n
a
tu
ra
l
lan
g
u
a
g
e
p
ro
c
e
ss
in
g
,
a
n
d
g
e
n
e
ra
ti
v
e
AI.
He
r
re
se
a
rc
h
fo
c
u
se
s
o
n
g
e
n
e
ra
ti
v
e
AI,
larg
e
lan
g
u
a
g
e
m
o
d
e
ls
(LL
M
s),
a
n
d
v
isu
a
l
lan
g
u
a
g
e
m
o
d
e
ls
(VL
M
s).
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
a
n
u
ra
d
h
a
@k
m
it
.
in
.
Dr
.
Ma
r
th
a
S
h
e
shi
k
a
l
a
is
c
u
rre
n
tl
y
h
o
l
d
in
g
th
e
p
o
siti
o
n
o
f
He
a
d
a
n
d
P
ro
fe
ss
o
r
a
t
th
e
S
c
h
o
o
l
o
f
Co
m
p
u
ter
S
c
ien
c
e
a
n
d
Artifi
c
ial
In
tell
ig
e
n
c
e
a
t
S
R
Un
iv
e
rsity
in
Wara
n
g
a
l
,
In
d
ia.
S
h
e
e
a
rn
e
d
h
e
r
P
h
.
D.
i
n
c
o
m
p
u
ter
sc
ien
c
e
a
n
d
e
n
g
i
n
e
e
rin
g
fro
m
K
L
Ed
u
c
a
ti
o
n
a
l
F
o
u
n
d
a
ti
o
n
,
A
n
d
h
ra
P
ra
d
e
sh
,
in
M
a
rc
h
2
0
1
8
.
Dr.
S
h
e
sh
ik
a
la'
s
re
se
a
rc
h
fo
c
u
se
s
o
n
a
re
a
s
su
c
h
a
s
d
a
ta
m
in
in
g
,
m
a
c
h
in
e
lea
rn
i
n
g
,
a
n
d
n
a
tu
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
.
He
r
e
x
ten
si
v
e
a
c
a
d
e
m
ic
c
o
n
tri
b
u
ti
o
n
s
in
c
l
u
d
e
o
v
e
r
5
0
p
u
b
li
c
a
ti
o
n
s
in
v
a
rio
u
s
n
a
ti
o
n
a
l
a
n
d
i
n
tern
a
ti
o
n
a
l
jo
u
rn
a
ls,
c
o
n
fe
re
n
c
e
s,
a
n
d
p
ro
c
e
e
d
i
n
g
s.
Sh
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
ma
rth
a
k
a
la0
8
@
g
m
a
il
.
c
o
m
.
Evaluation Warning : The document was created with Spire.PDF for Python.