I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
, pp.
289
~
299
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
15
.i
1
.pp
289
-
299
289
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
E
n
h
a
n
c
i
n
g
m
e
d
i
c
al
l
an
gu
age
m
od
e
l
s
w
i
t
h
b
i
g d
at
a t
e
c
h
n
ol
ogi
e
s
A
you
b
A
ll
al
i
1
, I
b
t
ih
al
A
b
o
u
c
h
ab
ak
a
2
, N
aj
a
t
R
af
al
ia
1
1
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
, F
a
c
ul
t
y of
S
c
i
e
nc
e
s
, I
bn T
of
a
i
l
U
ni
ve
r
s
i
t
y, K
e
ni
t
r
a
, M
or
oc
c
o
2
M
oha
m
m
e
d V
I
U
ni
ve
r
s
i
t
y of
S
c
i
e
nc
e
s
a
nd H
e
a
l
t
h,
C
a
s
a
bl
a
nc
a
,
M
or
oc
c
o
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
J
un
1
,
2025
R
e
vi
s
e
d
J
a
n
1
,
2026
A
c
c
e
pt
e
d
J
a
n
22
,
2026
In
this
study,
we
present
an
end
-
to
-
end,
big
-
data
–
driven
framewo
rk
for
continu
ously
enriching
and
fine
-
tuning
large
language
models
(LLM
s)
with
the
latest
professional
and
scientifi
c
medical
knowledge.
Streaming
u
pdates
from
premier
sources
such
as
The
New
England
Journal
of
Medicine
(NEJM)
are
ingested
via
an
Apache
Kafka
cluster
for
low
-
latency
d
elivery
and
durably
archived
in
a
three
-
node
Apache
Hadoop
(
Hadoop
distr
ibuted
file
system
(HDFS))
system.
Each
new
article
is
preprocessed
into
high
-
dimensional
embeddin
gs
and
indexed
in
a
Milvus
vector
database
to
enable
sub
-
second
semantic
retrieval
over
millions
of
records.
At
query
or
batch
time,
our
retrieval
-
augmented
generation
(RAG)
module
retrieves
th
e
top
-
k
relevant
embeddings
from
Milvus
and
injects
them
into
prompts
for
DeepSeek
-
R1,
GPT
-
4o
-
mini,
and
Llama
3,
models
which
are
hosted
,
fine
-
tuned,
and
served
via
Ollama
on
an
NVIDIA
GeForce
RTX
3050
T
i
GPU
for
efficient
inference
and
continual
learning.
The
enriched
outp
uts
are
seamlessly
delivered
to
end
u
sers
through
a
Telegram
bot
program
med
in
Python
using
the
Telebot
library,
linking
the
RAG
-
enhanced
LLMs
to
an
intuitive cha
t interfac
e. Our
Kafka,
HDFS, Milvus,
RAG, LLM,
or Tel
egram
bot
pipeline
demonstrably
improves
factual
accuracy
and
topical
curre
ncy
of
AI
-
generated
medical
insights
across
clinical
decision
support,
patient
engagement
and
education
,
drug
discovery
and
dev
elopment
,
virtual
health
assistan
ts,
and
mental
health
support
,
laying
the
groundwor
k
for
truly
intelligent, r
esponsive, a
nd data
-
d
riven healthcare s
olutions
.
K
e
y
w
o
r
d
s
:
B
ig
da
ta
D
e
e
pS
e
e
k
-
R1
G
P
T
4o
-
m
in
i
L
L
M
s
L
la
m
a
3
R
e
tr
ie
va
l
-
a
ugm
e
nt
e
d
ge
ne
r
a
ti
on
V
e
c
to
r
da
ta
ba
s
e
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
A
youb Alla
li
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
S
c
ie
nc
e
, F
a
c
ul
ty
of
S
c
ie
nc
e
s
, I
bn
T
of
a
il
U
ni
ve
r
s
it
y
K
e
ni
tr
a
, M
or
oc
c
o
E
m
a
il
:
a
youb.a
ll
a
li
@
ui
t.
a
c
.m
a
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
he
he
a
lt
hc
a
r
e
s
e
c
to
r
i
s
e
xpe
r
ie
n
c
in
g
a
pa
r
a
di
gm
s
hi
f
t
dr
iv
e
n
by
a
dva
nc
e
s
in
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
nd
bi
g
-
da
ta
te
c
hnol
ogi
e
s
.
I
n
pa
r
ti
c
ul
a
r
,
la
r
ge
l
a
ngua
ge
m
ode
ls
(
L
L
M
s
)
ha
ve
s
how
n
tr
e
m
e
ndous
pr
om
is
e
a
c
r
os
s
m
e
di
c
a
l
dom
a
in
s
,
f
r
om
c
li
ni
c
a
l
d
e
c
is
io
n
s
uppor
t
to
vi
r
tu
a
l
he
a
lt
h
a
s
s
i
s
ta
nt
s
.
B
ut
th
e
ir
r
e
a
l
-
w
or
ld
im
pa
c
t
de
pe
nds
on
a
c
c
e
s
s
to
hi
gh
-
qua
li
ty
,
up
-
to
-
da
te
knowle
dge
.
W
it
h
th
e
vol
um
e
of
m
e
di
c
a
l
li
te
r
a
tu
r
e
,
c
li
ni
c
a
l
s
tu
di
e
s
,
a
nd
e
xpe
r
t
c
om
m
e
nt
a
r
y
gr
ow
in
g
e
xpone
nt
ia
l
ly
,
it
r
e
m
a
in
s
a
pr
e
s
s
in
g
c
ha
ll
e
nge
to
in
ge
s
t,
s
to
r
e
,
pr
oc
e
s
s
,
a
nd
s
ur
f
a
c
e
th
is
e
v
e
r
-
e
xpa
ndi
ng
body
of
in
f
o
r
m
a
ti
on
in
w
a
ys
th
a
t
ke
e
p
pa
c
e
w
it
h
e
m
e
r
gi
ng
di
s
c
ove
r
ie
s
a
nd
e
vol
vi
ng s
ta
nda
r
ds
of
c
a
r
e
.
T
o
m
e
e
t
th
is
c
ha
ll
e
nge
,
w
e
in
tr
oduc
e
a
f
ul
ly
in
te
gr
a
te
d,
bi
g
-
da
ta
a
r
c
hi
te
c
tu
r
e
th
a
t
c
oupl
e
s
r
e
a
l
-
ti
m
e
s
tr
e
a
m
in
g,
s
c
a
la
bl
e
s
to
r
a
ge
,
s
e
m
a
nt
ic
in
d
e
xi
ng,
a
nd
r
e
tr
ie
va
l
-
a
ugm
e
nt
e
d
ge
ne
r
a
ti
on
(
R
A
G
)
to
c
ont
in
uous
ly
e
nr
ic
h
a
nd
f
in
e
-
tu
ne
L
L
M
s
f
or
m
e
di
c
a
l
a
ppl
ic
a
ti
ons
.
F
ir
s
t,
f
e
e
ds
f
r
om
le
a
di
ng
s
our
c
e
s
s
u
c
h
a
s
t
he
N
e
w
E
ngl
a
nd
J
our
na
l
of
M
e
di
c
in
e
(
N
E
J
M
)
a
nd
t
he
L
a
nc
e
t
a
r
e
c
a
pt
ur
e
d
vi
a
a
n
A
pa
c
he
K
a
f
ka
c
lu
s
te
r
,
e
ns
ur
in
g
s
ub
-
s
e
c
ond
de
li
ve
r
y
of
ne
w
ly
publ
is
h
e
d
a
r
ti
c
le
s
.
R
a
w
doc
u
m
e
nt
s
a
r
e
dur
a
bl
y
a
r
c
hi
ve
d
in
a
th
r
e
e
-
node
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
289
-
299
290
H
a
doop
di
s
tr
ib
ut
e
d
f
il
e
s
ys
te
m
(
H
D
F
S
)
f
or
f
a
ul
t
-
to
le
r
a
nt
,
di
s
tr
ib
ut
e
d
s
to
r
a
ge
.
E
a
c
h
in
c
om
in
g
r
e
c
or
d
is
th
e
n
pr
e
pr
oc
e
s
s
e
d
in
to
hi
gh
-
di
m
e
ns
io
na
l
e
m
be
ddi
ngs
a
nd
in
de
x
e
d
in
a
M
il
vus
ve
c
to
r
da
t
a
ba
s
e
,
e
na
bl
in
g
m
il
li
s
e
c
ond
-
s
c
a
le
s
e
m
a
nt
ic
r
e
tr
ie
va
l
a
c
r
os
s
m
il
li
ons
of
m
e
di
c
a
l
r
e
c
or
ds
.
A
t
que
r
y
or
ba
tc
h
ti
m
e
,
our
R
A
G
m
odul
e
r
e
tr
ie
ve
s
th
e
to
p
-
k
r
e
le
va
nt
ve
c
to
r
s
f
r
om
M
il
vus
a
nd
in
je
c
ts
th
e
ir
c
ont
e
nt
in
to
pr
om
pt
s
f
or
th
r
e
e
s
ta
te
-
of
-
th
e
-
a
r
t
L
L
M
s
,
D
e
e
pS
e
e
k
-
R
1,
G
P
T
-
4o
-
m
in
i,
a
nd
L
la
m
a
3
,
e
a
c
h
ho
s
te
d
a
nd
f
in
e
-
tu
ne
d
on
a
n
N
V
I
D
I
A
G
e
F
or
c
e
R
T
X
305
0
T
i
G
P
U
vi
a
O
ll
a
m
a
f
or
opt
im
iz
e
d
in
f
e
r
e
nc
e
a
nd
c
ont
in
ua
l
le
a
r
ni
ng.
F
in
a
ll
y,
c
li
ni
c
ia
ns
,
r
e
s
e
a
r
c
he
r
s
,
a
nd
pa
t
ie
nt
s
in
te
r
a
c
t
w
it
h
th
e
e
nr
ic
he
d
L
L
M
out
put
s
th
r
ough
a
P
yt
hon
-
ba
s
e
d
T
e
le
gr
a
m
bot
(
bui
lt
w
it
h
th
e
T
e
le
bot
li
br
a
r
y)
.
P
r
ovi
di
ng
a
n
in
tu
it
iv
e
c
ha
t
in
te
r
f
a
c
e
th
a
t
de
li
ve
r
s
e
vi
de
nc
e
-
ba
c
k
e
d i
ns
ig
ht
s
di
r
e
c
tl
y t
o e
nd u
s
e
r
s
'
m
o
bi
le
de
vi
c
e
s
.
T
he
pr
im
a
r
y
goa
l
of
th
is
r
e
s
e
a
r
c
h
is
to
de
m
ons
tr
a
te
th
e
p
ot
e
nt
ia
l
of
bi
g
da
ta
te
c
hnol
ogi
e
s
in
im
pr
ovi
ng
th
e
r
e
li
a
bi
li
ty
,
a
c
c
ur
a
c
y,
a
nd
r
e
a
l
-
ti
m
e
a
da
pt
a
bi
l
it
y
of
A
I
-
dr
iv
e
n
m
e
di
c
a
l
a
ppl
ic
a
ti
on
s
,
our
pr
opos
e
d f
r
a
m
e
w
or
k i
s
de
s
ig
ne
d t
o s
uppor
t
a
w
id
e
r
a
ng
e
of
he
a
lt
hc
a
r
e
us
e
c
a
s
e
s
, i
n
c
lu
di
ng:
i)
C
li
ni
c
a
l
de
c
is
io
n
s
uppor
t:
a
s
s
is
ti
ng
he
a
lt
hc
a
r
e
pr
of
e
s
s
io
na
ls
i
n
di
a
gnos
in
g
c
ondi
ti
ons
,
r
e
c
om
m
e
ndi
ng
tr
e
a
tm
e
nt
s
, a
nd i
de
nt
if
yi
ng pote
nt
ia
l
r
is
ks
ba
s
e
d on the
l
a
te
s
t
m
e
di
c
a
l
li
te
r
a
tu
r
e
.
ii)
P
a
ti
e
nt
e
nga
ge
m
e
nt
a
nd e
duc
a
ti
on:
pr
ovi
di
ng
pe
r
s
ona
li
z
e
d, e
vi
de
nc
e
-
ba
s
e
d
r
e
s
pons
e
s
to
pa
ti
e
nt
qu
e
r
ie
s
,
im
pr
ovi
ng he
a
lt
h l
it
e
r
a
c
y
,
a
nd s
e
lf
-
c
a
r
e
m
a
na
ge
m
e
nt
.
iii)
D
r
ug
di
s
c
ove
r
y
a
nd
de
ve
lo
pm
e
nt
:
a
c
c
e
le
r
a
ti
ng
pha
r
m
a
c
e
ut
ic
a
l
r
e
s
e
a
r
c
h
by
a
na
ly
z
in
g
va
s
t
da
ta
s
e
ts
of
c
li
ni
c
a
l
tr
ia
ls
, dr
ug i
nt
e
r
a
c
ti
ons
, a
nd biom
e
di
c
a
l
s
tu
di
e
s
.
iv
)
V
ir
tu
a
l
he
a
lt
h
a
s
s
i
s
ta
nt
s
:
e
nh
a
nc
in
g
te
le
m
e
di
c
in
e
s
e
r
vi
c
e
s
w
it
h
A
I
-
pow
e
r
e
d
c
ha
tb
ot
s
c
a
pa
bl
e
of
unde
r
s
ta
ndi
ng a
nd r
e
s
ponding t
o m
e
di
c
a
l
que
r
ie
s
w
it
h up
-
to
-
da
te
knowle
dge
.
v)
M
e
nt
a
l
he
a
lt
h
s
uppor
t:
le
ve
r
a
gi
ng
A
I
to
pr
ovi
de
c
onv
e
r
s
a
ti
on
a
l
s
uppor
t,
de
te
c
t
e
a
r
ly
s
ig
ns
of
m
e
nt
a
l
he
a
lt
h c
ondi
ti
ons
, a
nd r
e
c
om
m
e
nd a
ppr
opr
ia
te
i
nt
e
r
ve
nt
io
ns
.
B
ur
ga
n
e
t
al
.
[
1]
pr
e
s
e
nt
R
a
m
C
ha
t,
a
n
A
I
c
ha
tb
ot
de
s
ig
ne
d
t
o
he
lp
S
he
phe
r
d
U
ni
ve
r
s
it
y
s
tu
de
nt
s
na
vi
ga
te
th
e
ir
s
tu
de
nt
ha
ndbook,
de
v
e
lo
pe
d
in
P
yt
hon.
R
a
m
C
h
a
t
in
te
gr
a
te
s
bot
h
A
P
I
-
ba
s
e
d
a
nd
lo
c
a
l
L
L
M
s
us
in
g
th
e
L
a
ngC
ha
in
f
r
a
m
e
w
or
k
a
nd
a
ve
c
to
r
s
to
r
e
s
ys
te
m
.
T
he
c
ha
tb
ot
le
ve
r
a
ge
s
O
pe
nA
I
'
s
te
xt
-
e
m
be
ddi
ng
-
3
-
s
m
a
ll
m
ode
l
f
or
e
m
be
ddi
ngs
a
nd
in
it
ia
ll
y
us
e
d
O
p
e
nA
I
'
s
da
v
in
c
i
-
002
m
ode
l,
la
te
r
r
e
pl
a
c
e
d
w
it
h
ge
m
m
a
,
a
lo
c
a
l
L
L
M
ba
s
e
d on G
oogl
e
'
s
G
e
m
in
i
m
ode
l.
T
he
O
ll
a
m
a
f
r
a
m
e
w
or
k e
na
bl
e
s
a
ut
om
a
ti
c
L
L
M
s
e
le
c
ti
on ba
s
e
d
on
us
e
r
pr
om
pt
s
.
T
he
de
ve
lo
pm
e
nt
pr
oc
e
s
s
in
vol
ve
d
te
s
ti
ng
di
f
f
e
r
e
nt
L
L
M
s
,
de
bugging,
a
nd
opt
im
iz
in
g
R
a
m
C
ha
t'
s
pe
r
f
or
m
a
nc
e
.
T
he
ir
c
onf
e
r
e
nc
e
pr
e
s
e
nt
a
ti
on
w
il
l
c
o
ve
r
th
e
m
e
th
odol
ogy,
c
ha
ll
e
nge
s
,
a
nd
in
s
ig
ht
s
ga
in
e
d f
r
om
de
ve
lo
pi
ng t
hi
s
A
I
-
pow
e
r
e
d s
tu
de
nt
a
s
s
i
s
ta
nt
.
M
a
o
e
t
al
.
[
2]
di
s
c
us
s
th
e
c
ha
ll
e
ng
e
s
of
upda
ti
ng
L
L
M
s
w
it
h
lo
ng
-
ta
il
or
out
da
te
d
knowle
dge
due
to
th
e
ir
va
s
t
num
be
r
o
f
pa
r
a
m
e
te
r
s
,
m
a
ki
ng
f
in
e
-
tu
ni
ng
i
m
pr
a
c
ti
c
a
l.
I
ns
te
a
d,
th
e
y
hi
ghl
ig
ht
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
bl
a
c
k
-
box
R
A
G
,
w
hi
c
h
e
nha
nc
e
s
L
L
M
s
w
it
hout
m
odi
f
yi
ng
th
e
ir
pa
r
a
m
e
te
r
s
.
E
xi
s
ti
ng
bl
a
c
k
-
box
R
A
G
m
e
th
ods
of
te
n
f
in
e
-
tu
ne
th
e
r
e
tr
ie
ve
r
to
a
li
gn
w
it
h
L
L
M
pr
e
f
e
r
e
nc
e
s
but
f
a
c
e
two
ke
y
is
s
ue
s
:
ig
nor
in
g
f
a
c
tu
a
l
in
f
or
m
a
ti
on,
w
hi
c
h
c
a
n
m
is
le
a
d
th
e
r
e
tr
ie
ve
r
,
a
nd
in
e
f
f
ic
ie
nt
to
ke
n
us
a
ge
due
to
c
onc
a
te
na
ti
ng
a
ll
r
e
tr
ie
ve
d doc
um
e
nt
s
.
S
c
hi
e
le
e
t
al
.
[
3]
e
xa
m
in
e
th
e
im
pa
c
t
of
in
f
or
m
a
ti
on
a
nd
c
o
m
m
uni
c
a
ti
on
te
c
hnol
ogi
e
s
(
I
C
T
s
)
on
pol
it
ic
a
l
e
nga
ge
m
e
nt
,
pa
r
ti
c
ul
a
r
ly
in
th
e
c
ont
e
xt
of
vot
in
g
de
c
is
io
ns
,
w
it
h
th
e
r
is
e
of
is
s
ue
-
ba
s
e
d
vot
in
g
in
w
e
s
te
r
n
de
m
oc
r
a
c
ie
s
.
T
h
e
r
e
is
a
gr
ow
in
g
ne
e
d
f
or
tr
a
ns
pa
r
e
nt
a
nd
unbi
a
s
e
d
vot
in
g
a
dvi
c
e
a
ppl
ic
a
ti
ons
(
V
A
A
s
)
li
ke
S
w
it
z
e
r
la
nd'
s
S
m
a
r
tv
ot
e
a
nd
G
e
r
m
a
ny'
s
W
a
hl
-
O
-
M
a
t.
T
h
e
a
ut
hor
s
pr
opos
e
th
a
t
in
te
gr
a
ti
ng
L
L
M
s
w
it
h R
A
G
t
e
c
hni
que
s
c
oul
d e
nha
nc
e
V
A
A
s
by i
m
pr
ovi
n
g f
a
ir
ne
s
s
, i
m
pa
r
ti
a
li
ty
, a
nd t
r
a
ns
pa
r
e
nc
y.
W
hi
le
th
e
s
e
s
tu
di
e
s
de
m
ons
tr
a
te
a
dv
a
nc
e
m
e
nt
s
in
A
I
-
pow
e
r
e
d
a
ppl
ic
a
ti
ons
us
in
g
L
L
M
s
a
nd
R
A
G
,
th
e
y
la
r
ge
ly
ove
r
lo
ok
th
e
c
r
it
ic
a
l
r
ol
e
of
bi
g
da
ta
te
c
hnol
o
gi
e
s
in
e
n
s
ur
in
g
s
c
a
la
bi
li
ty
,
e
f
f
ic
ie
nc
y,
a
nd
r
e
a
l
-
ti
m
e
pr
oc
e
s
s
in
g.
T
he
ir
r
e
li
a
nc
e
on s
ta
ti
c
e
m
be
ddi
ngs
,
A
P
I
-
ba
s
e
d
m
ode
ls
, a
nd
f
in
e
-
tu
ne
d
r
e
tr
ie
ve
r
s
li
m
it
s
th
e
ir
a
bi
li
ty
to
ha
ndl
e
la
r
ge
-
s
c
a
le
,
c
ont
in
uous
ly
e
vol
vi
ng
d
a
ta
s
e
ts
.
U
nl
ik
e
th
e
s
e
a
ppr
oa
c
he
s
,
our
s
tu
dy
le
ve
r
a
ge
s
A
p
a
c
he
H
a
doop
f
or
di
s
tr
ib
ut
e
d
d
a
ta
s
to
r
a
ge
a
nd
A
pa
c
he
K
a
f
ka
f
or
r
e
a
l
-
ti
m
e
da
ta
s
tr
e
a
m
in
g,
e
na
bl
in
g dyna
m
ic
upda
te
s
, hi
gh
-
th
r
oughput pr
oc
e
s
s
in
g, a
nd i
m
pr
ove
d r
e
s
pons
iv
e
ne
s
s
. B
y i
nt
e
gr
a
ti
ng big da
ta
f
r
a
m
e
w
or
ks
,
our
r
e
s
e
a
r
c
h
a
ddr
e
s
s
e
s
ke
y
c
ha
ll
e
nge
s
in
la
r
ge
-
s
c
a
le
A
I
a
ppl
ic
a
ti
ons
th
a
t
th
e
s
e
s
tu
di
e
s
f
a
il
to
c
ons
id
e
r
, e
ns
ur
in
g a
m
or
e
s
c
a
la
bl
e
, a
nd da
t
a
-
dr
iv
e
n s
ol
ut
io
n.
T
r
a
di
ti
ona
l
L
L
M
s
li
ke
M
e
d
-
P
a
L
M
,
B
io
G
P
T
,
a
nd
P
ubM
e
d
G
P
T
a
r
e
ty
pi
c
a
ll
y
tr
a
in
e
d
on
s
ta
ti
c
da
ta
s
e
ts
,
l
e
a
di
ng
to
knowle
dge
c
ut
of
f
s
th
a
t
m
a
y
be
s
e
ve
r
a
l
m
on
th
s
or
e
ve
n
ye
a
r
s
ol
d. T
hi
s
s
ta
ti
c
n
a
tu
r
e
li
m
it
s
th
e
ir
a
bi
li
ty
to
in
c
or
por
a
te
a
nd
r
e
s
pond
to
th
e
la
te
s
t
m
e
di
c
a
l
r
e
s
e
a
r
c
h
pr
om
pt
ly
a
s
s
how
n
in
T
a
bl
e
1
[
4
]
.
I
n
c
ont
r
a
s
t,
R
A
G
s
y
s
te
m
s
e
nha
nc
e
L
L
M
s
by
in
te
gr
a
ti
ng
dyn
a
m
ic
r
e
tr
ie
va
l
m
e
c
h
a
ni
s
m
s
,
a
ll
ow
in
g
th
e
m
to
a
c
c
e
s
s
a
nd
ut
il
iz
e
up
-
to
-
da
te
in
f
or
m
a
ti
on
f
r
om
e
xt
e
r
na
l
da
ta
b
a
s
e
s
.
T
hi
s
a
ppr
oa
c
h
s
ig
ni
f
ic
a
nt
ly
r
e
duc
e
s
th
e
la
te
nc
y i
n i
nc
or
por
a
ti
ng ne
w
m
e
di
c
a
l
knowle
dge
i
nt
o t
he
m
ode
l
'
s
r
e
s
pons
e
s
.
T
hi
s
pa
pe
r
e
xpl
or
e
s
th
e
im
pl
e
m
e
nt
a
ti
on
de
ta
il
s
,
c
ha
ll
e
nge
s
,
a
nd
a
dva
nt
a
ge
s
of
us
in
g
bi
g
da
ta
pi
pe
li
ne
s
in
c
onj
unc
ti
on
w
it
h
R
A
G
-
e
nha
nc
e
d
L
L
M
s
f
or
m
e
di
c
a
l
a
ppl
ic
a
ti
ons
.
W
e
f
oc
us
on
th
e
im
pa
c
t
of
r
e
a
l
-
ti
m
e
da
ta
s
tr
e
a
m
in
g,
di
s
tr
ib
ut
e
d
s
to
r
a
ge
,
a
nd
f
r
e
que
nt
in
c
r
e
m
e
nt
a
l
tr
a
in
in
g,
w
hi
c
h
e
na
bl
e
our
m
ode
l
to
in
c
or
por
a
te
ne
w
ly
publ
is
he
d
m
e
di
c
a
l
r
e
s
e
a
r
c
h
on
a
da
il
y
ba
s
is
.
B
y
c
ont
r
a
s
t,
e
xi
s
ti
ng
m
ode
ls
s
uc
h
a
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g m
e
di
c
al
l
anguage
m
ode
ls
w
it
h bi
g data t
e
c
hnol
ogi
e
s
(
A
y
oub A
ll
al
i)
291
M
e
d
-
P
a
L
M
,
B
io
G
P
T
,
a
nd
P
ubM
e
dG
P
T
r
e
ly
on
s
ta
ti
c
pr
e
tr
a
in
i
ng
a
nd
r
e
qui
r
e
m
ont
hs
to
ye
a
r
s
to
upda
te
th
e
ir
knowle
dge
.
I
nt
e
gr
a
ti
ng
th
e
s
e
te
c
hnol
ogi
e
s
a
ll
ow
s
our
s
ys
te
m
to
pr
ovi
de
he
a
lt
hc
a
r
e
pr
of
e
s
s
io
na
ls
w
it
h
a
c
c
ur
a
te
,
ti
m
e
ly
,
a
nd
e
vi
de
nc
e
-
ba
s
e
d
r
e
s
pons
e
s
,
s
ig
ni
f
ic
a
nt
ly
im
pr
ovi
ng
th
e
pe
r
f
or
m
a
nc
e
,
r
e
s
pons
iv
e
ne
s
s
,
a
nd r
e
le
va
nc
e
of
A
I
-
dr
iv
e
n s
ol
ut
io
ns
i
n c
li
ni
c
a
l
de
c
i
s
io
n
-
m
a
ki
ng a
nd pa
ti
e
nt
c
a
r
e
.
T
a
bl
e
1. K
now
le
dg
e
upda
te
l
a
te
nc
y f
or
ne
w
r
e
s
e
a
r
c
h pa
pe
r
s
M
ode
l
K
now
l
e
dge
upda
t
e
m
e
t
hod
T
ypi
c
a
l
l
a
t
e
nc
y f
or
ne
w
pa
pe
r
M
e
d
-
P
a
L
M
/
M
e
d
-
P
a
L
M
-
2
S
t
a
t
i
c
pr
e
t
r
a
i
ne
d+f
i
ne
-
t
une
d
M
ont
hs
-
ye
a
r
s
B
i
oG
P
T
S
t
a
t
i
c
pr
e
t
r
a
i
ne
d m
ode
l
(
P
ubM
e
d s
na
ps
hot
)
M
ont
hs
-
ye
a
r
s
P
ubM
e
dG
P
T
S
t
a
t
i
c
pr
e
t
r
a
i
ne
d m
ode
l
(
P
ubM
e
d s
na
ps
hot
)
M
ont
hs
-
ye
a
r
s
2.
M
E
T
H
O
D
T
o
br
id
ge
bi
g
-
da
ta
te
c
hnol
ogi
e
s
w
it
h
L
L
M
s
in
a
m
e
di
c
a
l
s
e
tt
in
g,
w
e
de
s
ig
ne
d
a
f
our
-
s
ta
ge
pi
pe
li
ne
a
s
s
how
n
in
F
ig
ur
e
1.
D
a
ta
c
ol
le
c
ti
on
a
nd
s
tr
e
a
m
in
g,
w
he
r
e
pr
of
e
s
s
io
na
l
a
nd
s
c
i
e
nt
if
ic
m
e
di
c
a
l
ne
w
s
a
r
e
in
ge
s
te
d
in
r
e
a
l
-
ti
m
e
vi
a
A
pa
c
he
K
a
f
ka
.
D
i
s
tr
ib
ut
e
d
s
to
r
a
ge
,
w
hi
c
h
a
r
c
hi
ve
s
in
c
om
in
g
doc
um
e
nt
s
in
a
m
ul
ti
-
node
H
a
doop
H
D
F
S
c
lu
s
te
r
f
or
s
c
a
la
bl
e
,
f
a
ul
t
-
to
le
r
a
nt
pe
r
s
is
te
nc
e
.
S
e
m
a
nt
ic
e
nc
odi
ng
a
nd
R
A
G
in
te
gr
a
ti
on,
tr
a
ns
f
or
m
in
g
e
a
c
h
a
r
ti
c
le
in
to
hi
gh
-
di
m
e
ns
io
na
l
e
m
be
ddi
ngs
,
in
de
xi
ng
th
e
m
in
a
M
il
vus
ve
c
to
r
da
ta
ba
s
e
,
a
nd
a
ugm
e
nt
in
g
L
L
M
s
(
D
e
e
pS
e
e
k,
G
P
T
-
4o
-
m
in
i,
a
nd
L
la
m
a
3)
vi
a
R
A
G
.
I
nt
e
r
a
c
ti
ve
A
I
-
dr
iv
e
n
te
xt
ge
ne
r
a
ti
on,
de
li
ve
r
in
g
a
s
k/
a
ns
w
e
r
que
r
ie
s
a
nd
ge
ne
r
a
te
d
m
e
di
c
a
l
in
s
ig
ht
s
th
r
ough
a
us
e
r
-
f
a
c
in
g
in
te
r
f
a
c
e
(
T
e
le
gr
a
m
bot
)
[
5]
.
F
ig
ur
e
1.
P
r
opos
e
d
m
e
th
odol
ogy
T
o
c
r
e
a
te
a
c
om
pr
e
he
n
s
iv
e
a
nd
c
ont
in
uous
ly
upda
te
d
m
e
di
c
a
l
da
ta
s
e
t
f
or
e
nha
nc
in
g
L
L
M
s
,
w
e
im
pl
e
m
e
nt
e
d
a
n
a
ut
om
a
te
d
da
ta
c
ol
le
c
ti
on
pi
pe
li
ne
l
e
ve
r
a
gi
n
g
r
e
a
ll
y
s
im
pl
e
s
yndi
c
a
ti
on
(
R
S
S
)
f
e
e
ds
a
nd
A
pa
c
he
K
a
f
ka
.
W
e
f
oc
u
s
e
d
on
obt
a
in
in
g
m
e
di
c
a
l
li
te
r
a
tu
r
e
f
r
om
th
e
N
E
J
M
,
a
hi
ghl
y
r
e
put
a
bl
e
m
e
di
c
a
l
jo
ur
na
l,
by
ut
il
iz
in
g
it
s
publ
ic
ly
a
va
il
a
bl
e
R
S
S
f
e
e
ds
a
s
s
how
n
in
F
ig
ur
e
2.
T
hr
ough
th
e
s
e
f
e
e
ds
,
w
e
s
ys
te
m
a
ti
c
a
ll
y
m
oni
to
r
e
d
a
nd
r
e
tr
ie
ve
d
th
e
la
te
s
t
publ
is
he
d
a
r
t
ic
le
s
in
r
e
a
l
-
ti
m
e
.
E
a
c
h
ne
w
a
r
ti
c
le
id
e
nt
if
ie
d
vi
a
t
he
R
S
S
f
e
e
d w
a
s
a
ut
om
a
ti
c
a
ll
y downloa
de
d, pr
e
s
e
r
vi
ng i
ts
or
ig
in
a
l
la
yout
,
f
ig
u
r
e
s
, a
nd t
e
xt
ua
l
c
ont
e
nt
t
o
e
ns
ur
e
da
ta
i
nt
e
gr
it
y a
nd c
om
pl
e
te
ne
s
s
[
6]
.
T
he
r
e
a
l
-
ti
m
e
s
tr
e
a
m
in
g
c
om
pone
nt
w
a
s
m
a
na
g
e
d
us
in
g
A
pa
c
he
K
a
f
ka
,
c
onf
ig
ur
e
d
w
it
hi
n
our
th
r
e
e
-
node
H
a
doop
c
lu
s
te
r
(
one
N
a
m
e
N
ode
a
nd
two
D
a
ta
N
od
e
s
)
.
K
a
f
ka
'
s
pr
oduc
e
r
s
c
ont
in
uous
ly
m
oni
to
r
e
d
th
e
N
E
J
M
R
S
S
f
e
e
d
f
or
upda
te
s
,
a
ut
om
a
ti
c
a
ll
y
dow
nl
oa
di
ng
n
e
w
a
r
ti
c
le
s
a
s
s
oon
a
s
th
e
y
be
c
a
m
e
a
va
il
a
bl
e
.
U
pon
s
uc
c
e
s
s
f
ul
r
e
tr
ie
va
l,
K
a
f
ka
br
oke
r
s
di
s
tr
ib
ut
e
d
th
e
s
e
a
r
ti
c
le
s
a
c
r
os
s
de
s
ig
na
te
d
pa
r
ti
ti
ons
w
it
hi
n
th
e
K
a
f
ka
to
pi
c
s
,
f
a
c
il
it
a
ti
ng
ba
la
nc
e
d
lo
a
d
m
a
n
a
ge
m
e
nt
a
nd
e
ns
ur
in
g
f
a
ul
t
-
to
le
r
a
nc
e
dur
in
g
da
ta
in
ge
s
ti
on,
a
s
s
how
n i
n F
ig
ur
e
3
[
7]
–
[
9]
.
K
a
f
ka
c
ons
um
e
r
s
th
e
n
s
ys
te
m
a
ti
c
a
ll
y
pr
oc
e
s
s
e
d
th
e
s
e
a
r
ti
c
le
s
,
e
xt
r
a
c
ti
ng
r
e
le
va
nt
te
xt
u
a
l
c
ont
e
nt
th
r
ough
de
di
c
a
te
d
pa
r
s
in
g
m
e
c
h
a
ni
s
m
s
.
T
hi
s
e
xt
r
a
c
te
d
te
xt
w
a
s
s
tr
uc
tu
r
e
d
in
to
a
f
or
m
a
t
s
ui
ta
bl
e
f
or
s
ubs
e
que
nt
s
to
r
a
ge
in
th
e
H
D
F
S
.
T
he
in
te
gr
a
ti
on
of
K
a
f
ka
e
ns
ur
e
d
s
e
a
m
le
s
s
,
uni
nt
e
r
r
upt
e
d
s
tr
e
a
m
in
g
of
m
e
di
c
a
l
a
r
ti
c
le
s
di
r
e
c
tl
y
in
to
th
e
H
D
F
S
,
m
a
in
ta
in
in
g
a
n
u
p
-
to
-
da
te
,
s
tr
uc
tu
r
e
d,
a
nd
e
a
s
il
y
r
e
tr
ie
va
bl
e
r
e
pos
it
or
y, a
s
s
how
n i
n F
ig
ur
e
4.
T
hi
s
e
nh
a
nc
e
d
d
a
ta
c
ol
le
c
ti
on
m
e
th
od
a
ll
ow
e
d
u
s
to
a
c
h
ie
ve
r
obus
t,
r
e
a
l
-
ti
m
e
in
ge
s
ti
on
of
hi
gh
-
qua
li
ty
m
e
di
c
a
l
li
te
r
a
tu
r
e
,
la
yi
ng
a
s
tr
ong
f
ounda
ti
on
f
or
tr
a
in
in
g
a
nd
r
e
f
in
in
g
our
R
A
G
-
e
nha
nc
e
d
L
L
M
s
.
T
o
s
to
r
e
th
e
c
ont
in
uou
s
ly
c
ol
le
c
te
d
m
e
di
c
a
l
da
ta
,
w
e
de
pl
oye
d
a
H
a
doop
c
lu
s
te
r
w
it
h
th
r
e
e
node
s
,
th
e
H
D
F
S
is
us
e
d
to
pr
ovi
de
s
c
a
la
bl
e
,
f
a
ul
t
-
to
le
r
a
nt
s
to
r
a
ge
f
or
la
r
ge
vol
um
e
s
of
uns
tr
uc
tu
r
e
d
te
xt
da
ta
,
A
pa
c
he
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
289
-
299
292
K
a
f
ka
s
e
a
m
le
s
s
ly
in
te
gr
a
te
s
w
it
h
H
a
doop,
w
he
r
e
K
a
f
ka
c
ons
um
e
r
s
pul
l
th
e
s
tr
e
a
m
e
d
da
ta
a
nd
s
to
r
e
it
in
H
D
F
S
, t
he
H
a
doop c
lu
s
te
r
c
ons
i
s
ts
of
:
i)
O
ne
N
a
m
e
N
ode
:
m
a
na
g
e
s
t
he
m
e
t
a
da
ta
a
nd dir
e
c
to
r
y s
tr
uc
tu
r
e
of
t
he
di
s
tr
ib
ut
e
d f
il
e
s
ys
te
m
.
ii)
T
w
o
D
a
ta
N
od
e
s
:
s
to
r
e
th
e
a
c
tu
a
l
r
a
w
da
ta
f
il
e
s
a
nd
di
s
tr
ib
ut
e
th
e
s
to
r
a
ge
lo
a
d
f
or
r
e
dunda
nc
y
a
nd
hi
gh
a
va
il
a
bi
li
ty
.
F
ig
ur
e
2. T
he
N
E
J
M
R
S
S
f
e
e
d
F
ig
ur
e
3. A
pa
c
he
K
a
f
ka
pr
oduc
e
r
/c
ons
um
e
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g m
e
di
c
al
l
anguage
m
ode
ls
w
it
h bi
g data t
e
c
hnol
ogi
e
s
(
A
y
oub A
ll
al
i)
293
F
ig
ur
e
4. D
a
ta
e
xa
m
pl
e
D
a
ta
s
to
r
e
d
in
H
D
F
S
is
pr
e
pr
oc
e
s
s
e
d
a
nd s
tr
uc
tu
r
e
d
to
im
pr
ove
r
e
tr
ie
va
l
e
f
f
ic
ie
nc
y
dur
in
g
A
I
m
ode
l
tr
a
in
in
g
a
nd
in
f
e
r
e
nc
e
.
T
hi
s
a
r
c
hi
te
c
tu
r
e
a
ll
ow
s
e
f
f
ic
ie
nt
s
c
a
li
ng.
E
ns
ur
in
g
th
a
t
a
s
th
e
vol
um
e
of
m
e
di
c
a
l
li
te
r
a
tu
r
e
gr
ow
s
, t
he
s
ys
te
m
c
a
n h
a
ndl
e
i
nc
r
e
a
s
e
d da
ta
l
oa
d
s
w
it
hout
c
om
pr
om
is
in
g pe
r
f
or
m
a
nc
e
[
10]
–
[
13
]
.
I
n
our
pi
pe
li
ne
,
th
e
ve
c
to
r
da
ta
b
a
s
e
pl
a
ys
a
c
e
nt
r
a
l
r
ol
e
in
e
na
bl
in
g
f
a
s
t,
s
e
m
a
nt
i
c
a
ll
y
a
w
a
r
e
r
e
tr
ie
va
l
of
m
e
di
c
a
l
knowle
dge
.
A
f
te
r
in
ge
s
ti
ng
a
nd
a
r
c
hi
vi
ng
r
a
w
a
r
ti
c
l
e
s
,
e
a
c
h
doc
um
e
nt
is
pr
e
pr
oc
e
s
s
e
d,
to
ke
ni
z
e
d,
c
le
a
ne
d,
a
nd
p
a
s
s
e
d
th
r
ough
a
tr
a
ns
f
or
m
e
r
-
ba
s
e
d
e
n
c
ode
r
to
pr
oduc
e
f
ix
e
d
le
ngt
h
e
m
be
ddi
ng
ve
c
to
r
s
th
a
t
c
a
pt
ur
e
c
ont
e
xt
ua
l
m
e
a
ni
ng.
T
h
e
s
e
e
m
be
ddi
ng
s
a
r
e
th
e
n
in
ge
s
t
e
d
in
to
M
il
vus
.
W
e
c
onf
ig
ur
e
a
n
in
ve
r
te
d
f
il
e
w
it
h
pr
oduc
t
qua
nt
iz
a
ti
on
(
I
V
F
-
P
Q
)
in
de
x
to
ba
la
nc
e
s
e
a
r
c
h
s
pe
e
d
a
nd
m
e
m
or
y
f
oot
p
r
in
t,
a
nd
w
e
pe
r
io
di
c
a
ll
y r
e
bui
ld
i
nde
x s
ha
r
ds
t
o a
c
c
om
m
oda
te
ne
w
da
ta
w
it
hout
s
e
r
vi
c
e
i
nt
e
r
r
upt
io
n. A
t
que
r
y t
im
e
,
to
p
-
k
ne
a
r
e
s
t
ne
ig
hbor
s
e
a
r
c
he
s
r
e
tr
ie
ve
th
e
m
os
t
r
e
le
v
a
nt
a
r
ti
c
le
e
m
be
ddi
ngs
in
s
ub
-
s
e
c
ond
la
te
nc
y.
T
he
s
e
r
e
tr
ie
ve
d ve
c
to
r
s
a
r
e
t
he
n de
c
ode
d ba
c
k i
nt
o doc
um
e
nt
pa
s
s
a
ge
s
a
nd f
us
e
d i
nt
o pr
om
pt
s
f
or
our
R
A
G
m
odul
e
.
B
y
le
ve
r
a
gi
ng
M
il
vus
'
s
s
c
a
la
bl
e
a
r
c
hi
te
c
tu
r
e
a
nd
a
dva
n
c
e
d
in
de
xi
ng
te
c
hni
que
s
,
our
s
y
s
te
m
m
a
in
ta
in
s
m
il
li
s
e
c
ond
-
s
c
a
le
s
e
m
a
nt
ic
r
e
tr
ie
va
l
pe
r
f
or
m
a
nc
e
e
ve
n
a
s
th
e
m
e
di
c
a
l
c
or
pus
gr
ow
s
in
to
th
e
m
il
li
ons
of
r
e
c
or
ds
, e
ns
ur
in
g t
ha
t
L
L
M
s
a
lwa
ys
dr
a
w
on t
he
m
os
t
pe
r
ti
ne
nt
a
nd up
-
to
-
da
te
i
nf
or
m
a
ti
on
[
14
]
.
T
o
e
nha
n
c
e
th
e
c
a
p
a
bi
li
ti
e
s
of
L
L
M
s
in
ge
n
e
r
a
ti
ng
m
e
di
c
a
ll
y
a
c
c
ur
a
te
a
nd
c
ont
e
xt
ua
ll
y
r
e
le
va
nt
te
xt
,
w
e
im
pl
e
m
e
nt
R
A
G
,
in
s
te
a
d
of
s
ol
e
ly
r
e
ly
in
g
on
pr
e
-
tr
a
in
e
d
L
L
M
knowle
dge
,
R
A
G
e
na
bl
e
s
m
ode
ls
to
dyna
m
ic
a
ll
y
r
e
tr
ie
ve
r
e
le
va
nt
m
e
di
c
a
l
in
f
or
m
a
ti
on
f
r
om
th
e
H
D
F
S
-
s
to
r
e
d
da
ta
s
e
t
b
e
f
or
e
ge
ne
r
a
ti
ng
r
e
s
pons
e
s
[
15]
.
W
e
f
in
e
-
tu
ne
th
r
e
e
s
ta
te
-
of
-
th
e
-
a
r
t
L
L
M
s
(
D
e
e
pS
e
e
k
-
R
1,
G
P
T
-
4o
-
m
in
i,
a
nd
L
la
m
a
3)
us
in
g
th
e
c
ur
a
te
d m
e
di
c
a
l
da
ta
s
e
t
s
to
r
e
d i
n H
D
F
S
, t
he
f
in
e
-
tu
ni
ng pr
oc
e
s
s
i
nvol
ve
s
:
i)
D
a
ta
s
e
t
pr
e
pa
r
a
ti
on:
e
xt
r
a
c
ti
ng
ke
y m
e
di
c
a
l
a
r
ti
c
le
s
a
nd
s
tr
uc
tu
r
in
g t
he
m
f
or
f
in
e
-
tu
ni
ng.
ii)
M
ode
l
tr
a
in
in
g:
us
in
g
O
ll
a
m
a
,
a
n
ope
n
-
s
our
c
e
pl
a
tf
or
m
opt
im
iz
e
d
f
or
e
f
f
ic
ie
nt
L
L
M
de
pl
oym
e
nt
,
to
f
in
e
-
tu
ne
m
ode
ls
on
a
n
N
V
I
D
I
A
G
e
F
or
c
e
R
T
X
3050
T
i
G
P
U
,
th
is
s
e
tu
p
e
ns
ur
e
s
f
a
s
te
r
tr
a
in
in
g
ti
m
e
s
a
nd i
m
pr
ove
d i
nf
e
r
e
nc
e
pe
r
f
or
m
a
nc
e
.
iii)
R
A
G
im
pl
e
m
e
nt
a
ti
on:
c
om
bi
ni
ng
a
r
e
tr
ie
va
l
m
e
c
h
a
ni
s
m
w
it
h
L
L
M
s
,
w
he
r
e
m
ode
l
s
f
ir
s
t
s
e
a
r
c
h
th
e
H
D
F
S
-
s
to
r
e
d
da
ta
s
e
t
f
or
r
e
le
va
nt
in
f
or
m
a
ti
on
be
f
o
r
e
ge
ne
r
a
ti
ng
r
e
s
pons
e
s
,
th
is
r
e
duc
e
s
ha
ll
uc
in
a
ti
ons
a
nd e
nha
nc
e
s
t
he
f
a
c
tu
a
l
a
c
c
ur
a
c
y of
ge
ne
r
a
te
d m
e
di
c
a
l
in
s
ig
ht
s
.
T
o
r
ig
or
ous
ly
e
v
a
lu
a
te
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
our
r
e
tr
ie
va
l
pi
pe
li
n
e
,
w
e
e
m
pl
oye
d s
ta
nda
r
d
in
f
or
m
a
ti
on
r
e
tr
ie
va
l
m
e
tr
ic
s
,
na
m
e
ly
R
e
c
a
ll
@
k
a
nd
m
e
a
n
r
e
c
ip
r
oc
a
l
r
a
nk
(
M
R
R
)
.
R
e
c
a
ll
@
k
m
e
a
s
ur
e
s
th
e
pr
opor
ti
on
of
r
e
le
va
nt
doc
um
e
nt
s
c
a
pt
ur
e
d
w
it
hi
n
th
e
to
p
-
k
r
e
tr
ie
ve
d
r
e
s
ul
ts
,
r
e
f
le
c
ti
ng
th
e
br
e
a
dt
h
of
c
ov
e
r
a
ge
f
or
a
gi
ve
n
m
e
di
c
a
l
que
r
y. M
R
R
, i
n c
ont
r
a
s
t,
e
m
pha
s
iz
e
s
how
e
a
r
ly
t
he
f
ir
s
t
r
e
le
va
nt
doc
um
e
nt
a
ppe
a
r
s
i
n t
he
r
a
nke
d l
is
t,
r
e
w
a
r
di
ng
s
ys
te
m
s
th
a
t
s
ur
f
a
c
e
hi
ghl
y
pe
r
ti
ne
nt
in
f
or
m
a
ti
on
a
t
th
e
to
p
o
f
th
e
r
e
s
ul
ts
.
U
s
in
g
a
be
nc
hm
a
r
k
s
e
t
of
c
li
ni
c
a
l
a
nd
bi
om
e
di
c
a
l
que
s
ti
on
s
,
w
e
c
om
pa
r
e
d
our
da
il
y
upda
te
d
R
A
G
s
ys
t
e
m
a
ga
in
s
t
s
ta
ti
c
ba
s
e
li
ne
s
s
uc
h
a
s
M
e
d
-
P
a
L
M
,
B
io
G
P
T
,
a
nd
P
ubM
e
dG
P
T
.
T
he
r
e
s
ul
ts
de
m
ons
tr
a
te
th
a
t
our
a
ppr
oa
c
h
c
ons
is
te
nt
ly
a
c
hi
e
ve
s
hi
ghe
r
R
e
c
a
ll
@
k
a
nd
M
R
R
s
c
or
e
s
a
s
s
how
n
in
T
a
bl
e
2,
in
di
c
a
ti
ng
s
upe
r
io
r
c
ove
r
a
ge
of
ne
w
ly
publ
is
he
d
a
r
ti
c
le
s
a
nd
f
a
s
te
r
a
c
c
e
s
s
to
th
e
m
o
s
t
r
e
le
va
nt
e
vi
d
e
nc
e
.
T
hi
s
va
li
da
te
s
th
a
t
c
ont
in
uous
in
ge
s
ti
on
a
nd
in
de
xi
ng
of
m
e
di
c
a
l
li
te
r
a
tu
r
e
s
ubs
ta
nt
ia
ll
y
im
pr
ove
r
e
tr
ie
va
l
qua
li
ty
a
nd
di
r
e
c
tl
y
e
nha
nc
e
th
e
r
e
li
a
bi
li
t
y
of
dow
ns
tr
e
a
m
a
ns
w
e
r
ge
ne
r
a
ti
on.
T
a
bl
e
2.
E
va
lu
a
ti
on r
e
s
ul
ts
of
m
ode
ls
M
ode
l
R
e
c
a
l
l
@
10
R
e
c
a
l
l
@
20
M
R
R
@
10
M
R
R
@
20
O
ur
da
i
l
y R
A
G
m
o
de
l
(
G
P
T
4o
m
i
n
i
)
0.8
2
0.9
1
0.7
4
0.8
1
M
e
d
-
P
a
L
M
0.4
5
0.6
0
0.3
9
0.5
2
B
i
oG
P
T
0.4
8
0.6
2
0.4
2
0.5
5
P
ubM
e
dG
P
T
0.5
0
0.6
5
0.4
4
0.5
7
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
289
-
299
294
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
3.1.
AI
-
d
r
iv
e
n
t
e
xt
ge
n
e
r
at
io
n
f
or
m
e
d
ic
al
ap
p
li
c
at
io
n
s
O
nc
e
t
he
f
in
e
-
tu
ne
d m
ode
ls
a
r
e
opt
im
iz
e
d w
it
h R
A
G
, t
he
y a
r
e
de
pl
oye
d t
o ge
ne
r
a
te
m
e
di
c
a
l
in
s
ig
ht
s
in
va
r
io
us
a
ppl
ic
a
ti
ons
,
a
s
de
s
c
r
ib
e
d i
n t
he
f
ol
lo
w
in
g.
3.1.1.
C
li
n
ic
al
d
e
c
is
io
n
s
u
p
p
or
t
O
ur
R
A
G
-
e
nha
nc
e
d
L
L
M
s
c
ont
in
uous
ly
pul
l
in
th
e
la
te
s
t
p
e
e
r
-
r
e
vi
e
w
e
d
s
tu
di
e
s
a
nd
gui
de
li
ne
s
f
r
om
M
il
vus
,
s
ynt
he
s
iz
e
th
e
m
o
s
t
r
e
le
va
nt
f
in
di
ngs
,
a
nd
pr
e
s
e
nt
c
onc
is
e
,
e
vi
de
nc
e
-
ba
c
ke
d
s
um
m
a
r
ie
s
to
c
li
ni
c
ia
ns
.
B
y
e
m
be
ddi
ng
th
i
s
c
a
p
a
bi
li
ty
in
to
th
e
T
e
le
gr
a
m
bot
in
te
r
f
a
c
e
,
doc
to
r
s
c
a
n
que
r
y
c
om
pl
e
x
c
a
s
e
s
,
s
uc
h
a
s
e
m
e
r
gi
ng
th
e
r
a
pe
ut
ic
pr
ot
oc
ol
s
or
r
a
r
e
a
dve
r
s
e
e
ve
nt
s
,
a
n
d
r
e
c
e
iv
e
s
ubs
t
a
nt
ia
te
d
r
e
c
om
m
e
nda
ti
ons
in
s
e
c
onds
.
E
ns
ur
in
g
tr
e
a
tm
e
nt
de
c
is
io
ns
a
r
e
gr
ounde
d
in
th
e
m
os
t
c
ur
r
e
nt
m
e
di
c
a
l
li
te
r
a
tu
r
e
,
a
s
s
how
n
in
F
ig
ur
e
s
5 a
nd 6.
F
ig
ur
e
5. C
li
ni
c
a
l
de
c
is
io
n s
uppor
t
s
ys
t
e
m
pow
e
r
e
d by
G
P
T
4o mi
ni
w
it
h R
A
G
F
ig
ur
e
6. C
li
ni
c
a
l
de
c
is
io
n s
uppor
t
s
ys
t
e
m
pow
e
r
e
d by
B
io
G
P
T
5
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g m
e
di
c
al
l
anguage
m
ode
ls
w
it
h bi
g data t
e
c
hnol
ogi
e
s
(
A
y
oub A
ll
al
i)
295
3.1.2. P
at
ie
n
t
e
d
u
c
a
t
io
n
an
d
e
n
gage
m
e
n
t
T
he
s
y
s
te
m
le
ve
r
a
ge
s
th
e
s
a
m
e
s
e
m
a
nt
ic
r
e
tr
ie
va
l
pi
pe
li
ne
to
tr
a
ns
la
te
de
ns
e
c
li
ni
c
a
l
r
e
s
e
a
r
c
h
in
to
c
le
a
r
,
la
ype
r
s
on
-
f
r
ie
ndl
y
e
xpl
a
na
ti
ons
ta
il
or
e
d
to
in
di
vi
dua
l
pa
ti
e
nt
c
onc
e
r
ns
.
T
he
s
e
e
xpl
a
n
a
ti
ons
a
ddr
e
s
s
s
pe
c
if
ic
ne
e
d
s
,
w
h
e
th
e
r
m
e
di
c
a
ti
on s
id
e
e
f
f
e
c
ts
,
li
f
e
s
ty
le
m
odi
f
ic
a
ti
ons
,
or
pr
e
ve
nt
a
ti
ve
c
a
r
e
.
T
he
bot
de
li
ve
r
s
pe
r
s
ona
li
z
e
d,
up
-
to
-
da
te
gui
da
nc
e
th
a
t
e
m
pow
e
r
s
pa
ti
e
nt
s
to
be
tt
e
r
unde
r
s
ta
nd
th
e
ir
c
ondi
ti
ons
a
nd
a
dhe
r
e
to
tr
e
a
tm
e
nt
pl
a
ns
.
3.1.3. Dr
u
g d
is
c
ove
r
y an
d
d
e
ve
lo
p
m
e
n
t
P
ha
r
m
a
c
e
ut
ic
a
l
r
e
s
e
a
r
c
he
r
s
c
a
n
ha
r
ne
s
s
our
pl
a
tf
or
m
to
n
a
vi
ga
te
va
s
t
vol
um
e
s
of
tr
ia
l
da
ta
,
dr
ug
–
dr
ug
in
te
r
a
c
ti
on
r
e
por
ts
,
a
nd
m
ol
e
c
ul
a
r
s
tu
di
e
s
.
T
he
L
L
M
s
,
e
nr
ic
he
d
vi
a
R
A
G
,
hi
ghl
ig
ht
pr
om
is
in
g
c
om
pound
in
te
r
a
c
ti
ons
,
f
la
g
s
a
f
e
ty
s
ig
n
a
ls
,
a
nd
s
um
m
a
r
iz
e
c
li
ni
c
a
l
tr
ia
l
out
c
om
e
s
.
T
hi
s
dr
a
m
a
ti
c
a
ll
y
a
c
c
e
le
r
a
te
s
hypothe
s
is
g
e
ne
r
a
ti
on
a
nd
e
na
bl
e
s
m
or
e
in
f
or
m
e
d
de
c
is
io
ns
on
c
a
ndi
da
t
e
s
e
le
c
ti
on
a
nd
tr
ia
l
de
s
ig
n, a
s
s
how
n i
n F
ig
ur
e
7.
F
ig
ur
e
7. D
r
ug
di
s
c
ove
r
y a
nd de
ve
lo
pm
e
nt
by
G
P
T
4o mi
ni
w
it
h R
A
G
3.1.4.
V
ir
t
u
al
h
e
al
t
h
as
s
is
t
an
t
s
O
ur
P
yt
hon
-
dr
iv
e
n
T
e
le
gr
a
m
bot
s
e
r
ve
s
a
s
a
n
in
te
ll
ig
e
nt
f
r
ont
-
li
ne
a
id
e
,
e
nga
gi
ng
u
s
e
r
s
in
r
e
a
l
ti
m
e
to
tr
ia
ge
s
ym
pt
om
s
or
a
n
s
w
e
r
r
out
in
e
he
a
lt
h
in
qui
r
ie
s
.
B
y
c
o
upl
in
g
na
tu
r
a
l
-
la
ngua
ge
di
a
lo
gue
w
it
h
in
s
ta
nt
a
c
c
e
s
s
to
th
e
M
il
vus
-
in
de
xe
d
knowle
dge
ba
s
e
,
th
e
a
s
s
is
ta
n
t
ha
ndl
e
s
e
ve
r
yda
y
que
s
ti
ons
a
ut
onomous
ly
w
hi
le
e
s
c
a
l
a
ti
ng
c
r
it
ic
a
l
is
s
u
e
s
to
hum
a
n
pr
ovi
de
r
s
.
T
hi
s
a
ppr
oa
c
h
in
c
r
e
a
s
e
s
c
a
r
e
a
c
c
e
s
s
ib
il
it
y
a
nd
r
e
duc
e
s
c
li
ni
c
ia
n w
or
kl
oa
d.
3.1.5. M
e
n
t
al
h
e
al
t
h
s
u
p
p
or
t
T
hr
ough
e
m
pa
th
e
ti
c
c
onve
r
s
a
ti
ona
l
f
lo
w
s
pow
e
r
e
d
by
R
A
G
-
a
ugm
e
nt
e
d
L
L
M
s
,
th
e
s
ys
te
m
of
f
e
r
s
on
-
de
m
a
nd
m
e
nt
a
l
he
a
lt
h
c
he
c
k
-
in
s
,
c
opi
ng
s
tr
a
te
gi
e
s
r
oot
e
d
in
c
ogni
ti
ve
-
be
ha
vi
or
a
l
pr
in
c
ip
le
s
,
a
nd
e
a
r
ly
a
le
r
ts
f
or
c
onc
e
r
ni
ng
la
ngua
ge
pa
tt
e
r
ns
.
T
hi
s
a
lw
a
ys
-
a
va
il
a
bl
e
c
ha
t
in
te
r
f
a
c
e
pr
ovi
de
s
a
n
a
ddi
ti
ona
l
la
ye
r
of
s
uppor
t.
I
t
e
nc
our
a
ge
s
us
e
r
s
to
s
e
e
k
f
ur
th
e
r
c
a
r
e
w
he
n
ne
e
de
d,
w
hi
le
e
ns
ur
in
g
in
te
r
ve
nt
io
ns
a
r
e
in
f
or
m
e
d
by
th
e
l
a
te
s
t
ps
yc
hol
ogi
c
a
l
r
e
s
e
a
r
c
h
[
16]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
289
-
299
296
3.2. I
m
p
ac
t
of
r
e
al
-
t
im
e
d
at
a s
t
r
e
a
m
in
g an
d
d
is
t
r
ib
u
t
e
d
s
t
o
r
age
T
he
in
te
gr
a
ti
on
of
r
e
a
l
-
ti
m
e
da
ta
s
tr
e
a
m
in
g
a
nd
di
s
tr
ib
ut
e
d
s
to
r
a
ge
pl
a
ys
a
c
r
uc
ia
l
r
ol
e
in
e
ns
ur
in
g
th
e
e
f
f
ic
ie
nc
y,
s
c
a
la
bi
li
ty
,
a
nd
r
e
li
a
bi
li
ty
of
A
I
-
dr
iv
e
n
m
e
di
c
a
l
a
ppl
ic
a
ti
ons
.
A
pa
c
he
K
a
f
ka
,
a
s
a
r
e
a
l
-
ti
m
e
da
ta
s
tr
e
a
m
in
g
pl
a
tf
or
m
,
e
na
bl
e
s
c
ont
in
uous
in
ge
s
ti
on
of
pr
of
e
s
s
io
na
l
a
nd
s
c
i
e
nt
if
ic
m
e
di
c
a
l
n
e
w
s
f
r
om
tr
us
te
d
s
our
c
e
s
s
uc
h
a
s
M
e
d
s
c
a
p
e
,
th
e
N
E
J
M
,
a
nd
t
he
L
a
n
c
e
t.
B
y
le
v
e
r
a
gi
ng
K
a
f
ka
'
s
publ
is
h
-
s
ubs
c
r
ib
e
m
ode
l,
th
e
s
ys
te
m
e
n
s
ur
e
s
th
a
t
ne
w
ly
publ
is
he
d
m
e
di
c
a
l
r
e
s
e
a
r
c
h
is
pr
om
pt
ly
c
ol
le
c
te
d,
pr
oc
e
s
s
e
d,
a
nd
m
a
de
a
va
il
a
bl
e
f
or
A
I
m
ode
l
tr
a
in
in
g
a
nd
in
f
e
r
e
nc
e
.
T
hi
s
r
e
a
l
-
ti
m
e
in
ge
s
ti
on
c
a
pa
bi
li
ty
i
s
c
r
it
ic
a
l
in
th
e
m
e
di
c
a
l
dom
a
in
,
w
he
r
e
up
-
to
-
da
te
knowle
dge
is
e
s
s
e
nt
ia
l
f
or
a
c
c
ur
a
te
di
a
gnos
e
s
,
tr
e
a
tm
e
nt
r
e
c
om
m
e
nda
ti
ons
,
a
nd
pa
ti
e
nt
s
uppor
t.
A
ddi
ti
ona
ll
y,
H
a
doop
H
D
F
S
pr
ovi
de
s
a
r
obu
s
t,
di
s
tr
ib
ut
e
d
s
to
r
a
ge
s
ol
ut
io
n
th
a
t
e
f
f
ic
ie
nt
l
y
m
a
na
ge
s
l
a
r
ge
vol
um
e
s
of
m
e
di
c
a
l
te
xt
da
ta
w
hi
le
m
a
in
ta
in
in
g
hi
gh a
va
il
a
bi
li
ty
a
nd f
a
ul
t
to
le
r
a
nc
e
[
17]
.
T
hi
s
a
r
c
hi
te
c
tu
r
e
e
nh
a
nc
e
s
th
e
pe
r
f
or
m
a
nc
e
of
R
A
G
by
a
ll
ow
in
g
L
L
M
s
to
dyna
m
ic
a
ll
y
a
c
c
e
s
s
up
-
to
-
da
te
in
f
or
m
a
ti
on,
s
ig
ni
f
ic
a
nt
ly
r
e
duc
in
g
th
e
r
is
k
of
o
ut
da
te
d
or
in
c
or
r
e
c
t
A
I
-
ge
ne
r
a
te
d
r
e
s
pons
e
s
.
F
ur
th
e
r
m
or
e
,
th
e
us
e
of
a
di
s
tr
ib
ut
e
d
f
il
e
s
y
s
te
m
e
nha
n
c
e
s
s
c
a
la
bi
li
ty
,
e
ns
ur
in
g
th
a
t
th
e
s
ys
te
m
c
a
n
h
a
ndl
e
in
c
r
e
a
s
in
g
vol
um
e
s
of
m
e
di
c
a
l
li
te
r
a
tu
r
e
w
it
hout
de
gr
a
da
ti
o
n
in
pe
r
f
or
m
a
nc
e
.
T
he
c
om
bi
ne
d
pow
e
r
o
f
K
a
f
ka
'
s
r
e
a
l
-
ti
m
e
da
ta
in
ge
s
ti
on
a
nd
H
D
F
S
'
s
di
s
tr
ib
ut
e
d
s
to
r
a
g
e
e
na
bl
e
s
a
c
ont
in
uous
ly
e
vol
vi
ng
knowl
e
dge
ba
s
e
, i
m
pr
ovi
ng t
he
a
c
c
ur
a
c
y
a
nd r
e
le
va
nc
e
of
L
L
M
-
ge
ne
r
a
te
d
m
e
di
c
a
l
in
s
ig
ht
s
[
18]
.
3.3.
E
f
f
e
c
t
iv
e
n
e
s
s
o
f
R
A
G
i
n
r
e
d
u
c
in
g h
al
lu
c
in
at
io
n
s
T
he
in
te
gr
a
ti
on
of
R
A
G
s
ig
ni
f
ic
a
nt
ly
im
pr
ove
s
th
e
a
c
c
ur
a
c
y
a
nd
r
e
li
a
bi
li
ty
of
L
L
M
s
by
r
e
duc
in
g
ha
ll
uc
in
a
ti
ons
—
a
c
om
m
on
is
s
u
e
w
he
r
e
A
I
m
ode
ls
g
e
ne
r
a
te
m
is
le
a
di
ng
or
in
c
or
r
e
c
t
in
f
or
m
a
ti
on
due
to
li
m
it
a
ti
ons
in
th
e
ir
pr
e
-
tr
a
in
e
d
knowle
dge
.
T
r
a
di
ti
ona
l
L
L
M
s
r
e
ly
s
ol
e
ly
on
pr
e
-
e
xi
s
ti
ng
tr
a
in
in
g
da
ta
,
w
hi
c
h
c
a
n
be
c
om
e
out
da
te
d
or
la
c
k
dom
a
in
-
s
pe
c
if
ic
de
ta
il
s
,
pa
r
ti
c
ul
a
r
ly
in
dyna
m
ic
f
ie
ld
s
li
ke
m
e
di
c
in
e
.
R
A
G
m
it
ig
a
te
s
th
is
by
in
c
or
por
a
ti
ng
a
r
e
tr
ie
va
l
m
e
c
ha
ni
s
m
th
a
t
a
c
c
e
s
s
e
s
th
e
m
os
t
r
e
c
e
nt
,
c
ont
e
xt
ua
ll
y
r
e
le
va
nt
m
e
di
c
a
l
li
te
r
a
tu
r
e
f
r
om
H
a
doop
H
D
F
S
,
pow
e
r
e
d
by
r
e
a
l
-
ti
m
e
da
ta
s
tr
e
a
m
e
d
f
r
om
A
pa
c
he
K
a
f
ka
,
e
n
s
ur
in
g
f
a
c
tu
a
ll
y gr
ounde
d a
nd up
-
to
-
da
te
r
e
s
pons
e
s
.
I
n
our
im
pl
e
m
e
nt
a
ti
on,
a
ll
th
r
e
e
L
L
M
s
(
D
e
e
pS
e
e
k
-
R
1,
G
P
T
-
4o
-
m
in
i,
a
nd
L
la
m
a
3)
de
m
ons
tr
a
te
d
im
pr
ove
d
a
c
c
ur
a
c
y
a
nd
c
ont
e
xt
u
a
l
r
e
le
va
nc
e
w
he
n
e
nha
n
c
e
d
w
it
h
R
A
G
,
pa
r
ti
c
ul
a
r
ly
in
c
li
ni
c
a
l
de
c
is
io
n
s
uppor
t,
dr
ug
di
s
c
ove
r
y,
a
nd
pa
ti
e
nt
e
duc
a
ti
on
us
e
c
a
s
e
s
.
T
he
m
ode
ls
w
e
r
e
a
bl
e
to
r
e
f
e
r
e
nc
e
th
e
la
te
s
t
m
e
di
c
a
l
s
tu
di
e
s
a
nd
gui
de
li
ne
s
,
s
ig
ni
f
ic
a
nt
ly
r
e
duc
in
g
s
pe
c
ul
a
ti
ve
or
in
c
or
r
e
c
t
r
e
s
pons
e
s
.
A
ddi
ti
ona
ll
y,
R
A
G
e
nha
nc
e
s
e
xpl
a
in
a
bi
li
ty
by
a
ll
ow
in
g
m
ode
l
s
to
c
it
e
s
p
e
c
if
ic
doc
um
e
nt
s
or
s
our
c
e
s
,
th
e
r
e
by
im
pr
ovi
ng
tr
us
twor
th
in
e
s
s
i
n c
r
it
ic
a
l
m
e
di
c
a
l
a
ppl
ic
a
ti
on
s
[
19]
.
3.4.
P
e
r
f
or
m
an
c
e
ac
r
o
s
s
d
i
f
f
e
r
e
n
t
L
L
M
s
T
he
pe
r
f
or
m
a
nc
e
of
th
e
th
r
e
e
f
in
e
-
tu
ne
d
L
L
M
s
(
D
e
e
pS
e
e
k
-
R
1,
G
P
T
-
4o
-
m
in
i,
a
nd
L
la
m
a
3)
va
r
ie
d
a
c
r
os
s
di
f
f
e
r
e
nt
m
e
di
c
a
l
a
ppl
ic
a
ti
ons
,
hi
ghl
ig
ht
in
g
th
e
s
tr
e
ngt
hs
a
nd
tr
a
de
-
of
f
s
of
e
a
c
h
m
ode
l
in
ha
ndl
in
g
c
om
pl
e
x
he
a
lt
hc
a
r
e
-
r
e
la
te
d
que
r
ie
s
.
G
P
T
-
4o
-
m
in
i
de
m
ons
tr
a
te
d
e
xc
e
pt
io
na
l
na
tu
r
a
l
la
ngu
a
ge
f
lu
e
nc
y,
m
a
ki
ng
it
pa
r
ti
c
ul
a
r
ly
e
f
f
e
c
ti
ve
in
-
pa
ti
e
nt
e
nga
ge
m
e
nt
,
m
e
nt
a
l
he
a
lt
h
s
uppor
t,
a
nd
vi
r
tu
a
l
he
a
lt
h
a
s
s
is
t
a
nt
a
ppl
ic
a
ti
ons
w
he
r
e
c
onve
r
s
a
ti
ona
l
c
ohe
r
e
nc
e
a
nd
e
m
ot
io
na
l
in
te
ll
ig
e
nc
e
a
r
e
c
r
uc
ia
l,
it
s
a
bi
li
ty
to
ge
ne
r
a
te
c
ont
e
xt
-
a
w
a
r
e
r
e
s
pons
e
s
w
it
h
hi
gh
r
e
a
da
bi
li
ty
m
a
de
it
th
e
pr
e
f
e
r
r
e
d
c
hoi
c
e
f
or
s
c
e
na
r
io
s
r
e
qui
r
in
g
hum
a
n
-
li
ke
in
te
r
a
c
ti
on
a
nd
e
m
pa
th
e
ti
c
c
om
m
uni
c
a
ti
on. De
e
pS
e
e
k
-
R
1,
on
th
e
ot
he
r
ha
nd, e
xc
e
ll
e
d
in
m
e
di
c
a
l
r
e
s
e
a
r
c
h
-
or
ie
nt
e
d
ta
s
k
s
,
s
uc
h
a
s
s
um
m
a
r
iz
in
g
c
li
ni
c
a
l
tr
ia
l
da
ta
,
e
xt
r
a
c
ti
ng
ke
y
f
in
di
ngs
f
r
om
m
e
di
c
a
l
li
te
r
a
tu
r
e
,
a
nd
id
e
nt
if
yi
ng
pot
e
nt
ia
l
dr
ug
in
te
r
a
c
ti
ons
,
it
s
s
tr
ong
a
na
ly
ti
c
a
l
c
a
pa
bi
li
ti
e
s
a
ll
ow
e
d
f
or
m
or
e
s
tr
uc
tu
r
e
d,
in
f
or
m
a
ti
on
-
de
ns
e
out
put
s
,
m
a
ki
ng
it
hi
ghl
y
s
ui
ta
bl
e
f
or
s
c
ie
nt
if
ic
a
nd
pha
r
m
a
c
e
ut
ic
a
l
r
e
s
e
a
r
c
h
a
ppl
ic
a
ti
ons
.
L
la
m
a
3
pr
ovi
de
d
a
ba
la
nc
e
d
pe
r
f
or
m
a
nc
e
a
c
r
os
s
m
ul
ti
pl
e
us
e
c
a
s
e
s
,
de
m
on
s
tr
a
ti
ng
r
obus
t
c
ont
e
xt
ua
l
unde
r
s
ta
ndi
ng
in
c
li
ni
c
a
l
d
e
c
is
io
n
s
uppor
t
w
hi
le
m
a
in
ta
in
in
g
r
e
a
s
ona
bl
e
f
lu
e
nc
y
in
pa
ti
e
nt
-
or
ie
nt
e
d di
a
lo
gue
s
, i
ts
e
f
f
ic
ie
nc
y i
n r
e
a
l
-
ti
m
e
R
A
G
w
or
kf
lo
w
s
e
ns
ur
e
d t
ha
t
m
e
di
c
a
l
in
s
ig
ht
s
w
e
r
e
c
ons
is
t
e
nt
ly
a
c
c
ur
a
te
a
nd w
e
ll
-
r
e
f
e
r
e
nc
e
d.
I
nt
e
gr
a
ti
ng
th
e
s
e
m
ode
ls
w
it
h
A
pa
c
he
K
a
f
ka
a
nd
H
a
doop
H
D
F
S
e
na
bl
e
d
c
ont
in
uous
upda
te
s
a
nd
f
in
e
-
tu
ni
ng
us
in
g
f
r
e
s
h
m
e
di
c
a
l
da
ta
,
r
e
duc
in
g
r
e
li
a
nc
e
on
s
ta
ti
c
knowle
dge
ba
s
e
s
.
H
ow
e
ve
r
,
m
ode
l
pe
r
f
or
m
a
nc
e
w
a
s
in
f
lu
e
nc
e
d
by
c
om
put
a
ti
ona
l
c
on
s
tr
a
in
ts
,
w
it
h
G
P
T
-
4o
-
m
in
i
c
ons
um
in
g
m
or
e
r
e
s
our
c
e
s
due
to
it
s
a
dva
nc
e
d
r
e
a
s
oni
ng
c
a
pa
bi
li
ti
e
s
,
w
hi
le
D
e
e
pS
e
e
k
-
R
1
a
n
d
L
la
m
a
3
of
f
e
r
e
d
be
tt
e
r
e
f
f
ic
ie
nc
y
-
a
c
c
ur
a
c
y
ba
la
nc
e
.
A
ddi
ti
ona
ll
y,
ha
ndl
in
g
a
m
bi
guous
m
e
di
c
a
l
que
r
ie
s
po
s
e
d
c
ha
ll
e
nge
s
,
a
s
di
f
f
e
r
e
nc
e
s
in
e
a
c
h
m
ode
l'
s
tr
a
in
in
g a
r
c
hi
te
c
tu
r
e
a
f
f
e
c
te
d doc
um
e
nt
pr
io
r
it
iz
a
ti
on f
r
om
H
D
F
S
[
20]
, [
21
]
.
3.5. Ch
al
le
n
ge
s
an
d
l
im
it
at
io
n
s
D
e
s
pi
te
th
e
pr
om
is
in
g
r
e
s
ul
ts
of
in
te
gr
a
ti
ng
bi
g
da
ta
te
c
hn
ol
ogi
e
s
w
it
h
R
A
G
-
e
nha
nc
e
d
L
L
M
s
,
s
e
ve
r
a
l
c
ha
ll
e
nge
s
a
nd l
im
it
a
ti
ons
m
us
t
be
a
ddr
e
s
s
e
d t
o opti
m
iz
e
t
he
ir
r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
on i
n t
he
he
a
lt
hc
a
r
e
dom
a
in
.
O
ne
of
th
e
pr
im
a
r
y
c
ha
ll
e
nge
s
is
c
om
put
a
ti
ona
l
r
e
s
o
ur
c
e
c
ons
tr
a
in
ts
,
a
s
f
in
e
-
tu
ni
ng
a
nd
de
pl
oyi
ng
L
L
M
s
on l
a
r
ge
-
s
c
a
le
m
e
di
c
a
l
da
ta
s
e
ts
r
e
qui
r
e
s
ig
ni
f
ic
a
nt
G
P
U
pow
e
r
a
nd me
m
or
y. W
hi
le
ou
r
s
ys
te
m
ut
il
iz
e
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
E
nhanc
in
g m
e
di
c
al
l
anguage
m
ode
ls
w
it
h bi
g data t
e
c
hnol
ogi
e
s
(
A
y
oub A
ll
al
i)
297
a
n
N
V
I
D
I
A
G
e
F
or
c
e
R
T
X
3050
T
i,
tr
a
in
in
g
la
r
ge
r
m
ode
ls
on
hi
gh
-
di
m
e
ns
io
na
l
m
e
di
c
a
l
da
ta
r
e
m
a
in
s
c
om
put
a
ti
ona
ll
y
e
xpe
ns
iv
e
a
nd
ti
m
e
-
in
te
ns
iv
e
,
li
m
it
in
g
th
e
f
e
a
s
ib
il
it
y
of
r
e
a
l
-
ti
m
e
f
in
e
-
tu
ni
ng
f
or
s
m
a
ll
e
r
he
a
lt
hc
a
r
e
or
ga
ni
z
a
ti
ons
.
D
a
ta
qua
li
ty
a
nd
bi
a
s
pos
e
c
r
it
ic
a
l
c
onc
e
r
ns
,
a
s
A
I
m
ode
ls
de
pe
nd
on
th
e
r
e
li
a
bi
li
t
y
of
th
e
ir
tr
a
in
in
g
a
nd
r
e
tr
ie
va
l
da
ta
.
E
ve
n
th
ough
m
e
di
c
a
l
li
te
r
a
tu
r
e
f
r
om
N
E
J
M
w
a
s
us
e
d
a
s
a
pr
im
a
r
y
da
t
a
s
our
c
e
, i
nhe
r
e
nt
bi
a
s
e
s
i
n c
li
ni
c
a
l
r
e
s
e
a
r
c
h, di
s
pa
r
it
ie
s
i
n pa
ti
e
nt
de
m
ogr
a
phi
c
s
, a
nd outda
te
d m
e
di
c
a
l
f
in
di
ngs
c
oul
d pote
nt
ia
ll
y s
ke
w
m
ode
l
pr
e
di
c
ti
ons
[
22]
, [
23]
.
E
ns
ur
in
g
da
ta
di
ve
r
s
it
y,
de
-
bi
a
s
in
g
m
e
th
odol
ogi
e
s
,
a
nd
c
ont
in
uous
va
li
da
ti
on
by
m
e
di
c
a
l
pr
of
e
s
s
io
na
ls
is
e
s
s
e
nt
ia
l
to
m
it
ig
a
te
th
e
s
e
r
is
k
s
,
a
not
he
r
li
m
it
a
ti
on
is
th
e
r
e
tr
ie
va
l
la
te
nc
y
w
it
hi
n
th
e
A
pa
c
h
e
H
a
doop
H
D
F
S
e
c
os
ys
te
m
,
pa
r
ti
c
ul
a
r
ly
w
he
n
ha
ndl
in
g
la
r
g
e
-
s
c
a
le
uns
tr
uc
tu
r
e
d
te
xt
da
ta
,
w
hi
le
R
A
G
im
pr
ove
s
f
a
c
tu
a
l
a
c
c
ur
a
c
y,
in
e
f
f
ic
ie
nt
r
e
tr
ie
va
l
m
e
c
ha
ni
s
m
s
c
oul
d
s
lo
w
dow
n
r
e
s
pons
e
ti
m
e
s
,
a
f
f
e
c
ti
ng
us
a
bi
li
ty
in
ti
m
e
-
s
e
ns
it
iv
e
a
ppl
ic
a
ti
ons
li
ke
c
li
ni
c
a
l
de
c
is
io
n
s
u
ppor
t.
R
e
gul
a
to
r
y
a
nd
e
th
ic
a
l
c
onc
e
r
ns
r
e
m
a
in
s
ig
ni
f
ic
a
nt
ba
r
r
ie
r
s
to
de
pl
oym
e
nt
,
a
s
A
I
-
ge
ne
r
a
te
d
m
e
di
c
a
l
in
s
ig
ht
s
m
us
t
c
om
pl
y
w
it
h
he
a
lt
hc
a
r
e
r
e
gul
a
ti
ons
s
uc
h
a
s
he
a
lt
h
in
s
ur
a
nc
e
por
ta
bi
li
ty
a
nd
a
c
c
ount
a
bi
li
ty
a
c
t
(
H
I
P
A
A
)
a
nd
ge
ne
r
a
l
da
ta
p
r
ot
e
c
ti
on
r
e
gul
a
ti
on
(
G
D
P
R
)
to
e
ns
ur
e
da
ta
pr
iv
a
c
y
a
nd
s
e
c
ur
it
y,
th
e
r
is
k
of
m
is
in
te
r
pr
e
ta
ti
on
a
nd
ove
r
-
r
e
li
a
nc
e
on
A
I
-
ge
ne
r
a
te
d
r
e
c
om
m
e
nda
ti
ons
a
ls
o
hi
ghl
ig
ht
s
th
e
ne
e
d
f
or
e
xpl
a
in
a
bi
li
ty
a
nd
in
te
r
pr
e
ta
bi
li
ty
f
r
a
m
e
w
or
ks
,
a
ll
ow
in
g
doc
to
r
s
to
ve
r
if
y
A
I
s
ugge
s
ti
ons
be
f
or
e
m
a
ki
ng
c
r
it
ic
a
l
m
e
di
c
a
l
de
c
is
io
ns
.
L
a
s
tl
y,
ha
ndl
in
g
a
m
bi
guous
m
e
di
c
a
l
que
r
ie
s
r
e
m
a
in
s
a
c
ha
ll
e
nge
,
a
s
L
L
M
s
m
a
y
s
tr
uggl
e
w
it
h
va
gue
s
ym
pt
om
s
,
r
a
r
e
di
s
e
a
s
e
s
,
or
c
onf
li
c
ti
ng
m
e
di
c
a
l
opi
ni
ons
,
f
ut
ur
e
im
pr
ove
m
e
nt
s
s
houl
d
f
oc
us
on
hybr
id
r
e
tr
ie
va
l
m
ode
l
s
c
om
bi
ni
ng
ke
yw
or
d
-
ba
s
e
d
a
nd
s
e
m
a
nt
ic
s
e
a
r
c
h
te
c
hni
que
s
,
opt
im
iz
e
d
i
nde
xi
ng
s
tr
a
te
gi
e
s
f
or
f
a
s
te
r
a
c
c
e
s
s
to
s
to
r
e
d
m
e
di
c
a
l
da
ta
,
a
nd
c
ol
la
bor
a
ti
ve
A
I
-
hum
a
n
de
c
is
io
n
-
m
a
ki
ng
f
r
a
m
e
w
or
ks
to
m
a
xi
m
iz
e
r
e
li
a
bi
li
ty
,
a
ddr
e
s
s
in
g
th
e
s
e
c
ha
ll
e
ng
e
s
w
il
l
be
c
r
uc
ia
l
in
e
nh
a
nc
in
g
th
e
s
c
a
la
bi
li
ty
,
a
c
c
ur
a
c
y,
a
nd
tr
us
twor
th
in
e
s
s
of
A
I
-
dr
iv
e
n
he
a
lt
hc
a
r
e
s
ol
ut
io
ns
i
n c
li
ni
c
a
l
pr
a
c
ti
c
e
[
24]
–
[
26]
.
4.
C
O
N
C
L
U
S
I
O
N
T
hi
s
r
e
s
e
a
r
c
h
de
m
ons
tr
a
te
s
th
a
t
in
te
gr
a
ti
ng
s
c
a
la
bl
e
bi
g
da
ta
in
f
r
a
s
tr
uc
tu
r
e
s
w
it
h
R
A
G
,
e
nha
n
c
e
d
L
L
M
s
c
a
n dr
a
m
a
ti
c
a
ll
y i
m
pr
ove
t
he
r
e
le
va
nc
e
, a
c
c
ur
a
c
y, a
nd t
im
e
li
ne
s
s
of
A
I
-
dr
iv
e
n m
e
di
c
a
l
a
ppl
ic
a
ti
ons
by
s
tr
e
a
m
in
g
a
nd
a
r
c
hi
vi
ng
pr
of
e
s
s
io
na
l
m
e
di
c
a
l
ne
w
s
,
s
e
m
a
nt
ic
a
l
ly
in
de
xi
ng
m
il
li
ons
of
doc
um
e
nt
s
in
M
il
vus
,
a
nd
f
in
e
-
tu
ni
ng
s
ta
te
-
of
-
th
e
-
a
r
t
L
L
M
s
f
or
s
ub
s
e
c
ond,
e
vi
de
nc
e
-
ba
c
ke
d
in
s
ig
ht
s
.
T
o
a
dva
nc
e
th
is
f
r
a
m
e
w
or
k,
f
ut
ur
e
w
or
k
m
us
t
dr
iv
e
ul
tr
a
-
lo
w
-
la
te
nc
y
r
e
tr
ie
va
l
th
r
ough
o
pt
im
iz
e
d
ve
c
to
r
in
de
xi
ng
a
nd
hybr
id
s
e
a
r
c
h
s
tr
a
te
gi
e
s
,
e
xt
e
nd
m
ode
l
c
a
pa
bi
li
ti
e
s
by
f
us
in
g
m
ul
ti
-
m
oda
l
da
ta
s
uc
h
a
s
r
a
di
ol
ogy
im
a
ge
s
,
e
le
c
tr
oni
c
he
a
lt
h
r
e
c
or
ds
,
a
nd
ge
nom
ic
s
in
to
a
uni
f
ie
d
e
m
be
ddi
ng
s
pa
c
e
,
a
nd
e
m
be
d
e
xpl
a
in
a
bi
li
ty
vi
a
e
xpl
a
in
a
bl
e
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
X
A
I
)
m
odul
e
s
th
a
t
tr
a
c
e
e
a
c
h
r
e
c
om
m
e
nda
ti
on
ba
c
k
to
it
s
s
our
c
e
.
E
qua
ll
y
c
r
it
ic
a
l
is
e
ns
ur
in
g
e
th
ic
a
l,
c
om
pl
ia
nt
de
pl
oym
e
nt
,
im
pl
e
m
e
nt
in
g
pr
iv
a
c
y
s
a
f
e
gua
r
ds
,
hum
a
n
-
in
-
th
e
-
lo
op
ve
r
if
ic
a
ti
on,
a
nd
a
dhe
r
e
nc
e
to
H
I
P
A
A
,
G
D
P
R
,
a
nd e
m
e
r
gi
ng
A
I
r
e
gul
a
ti
ons
,
to
m
it
ig
a
te
bi
a
s
a
nd
ove
r
-
r
e
li
a
nc
e
on
a
ut
om
a
ti
on;
f
in
a
ll
y,
de
m
oc
r
a
ti
z
in
g
a
c
c
e
s
s
a
nd
s
uppor
ti
ng
c
ont
in
ua
l
le
a
r
n
in
g
a
c
r
os
s
in
s
ti
tu
ti
ons
w
il
l
r
e
qui
r
e
s
c
a
la
bl
e
,
di
s
tr
ib
ut
e
d
tr
a
in
in
g
a
ppr
oa
c
he
s
,
in
c
lu
di
ng
c
lo
ud
-
ba
s
e
d
pl
a
tf
o
r
m
s
,
f
e
de
r
a
te
d
le
a
r
ni
ng,
a
nd
e
dge
-
di
s
tr
ib
u
te
d
G
P
U
c
lu
s
te
r
s
.
B
y
a
ddr
e
s
s
in
g
th
e
s
e
c
h
a
ll
e
nge
s
,
w
e
c
a
n
tr
a
n
s
f
or
m
to
da
y'
s
pr
oof
-
of
-
c
onc
e
pt
in
to
a
gl
oba
ll
y
de
pl
oya
bl
e
,
r
e
a
l
-
ti
m
e
,
a
nd
tr
us
twor
th
y
A
I
-
pow
e
r
e
d
m
e
di
c
a
l
de
c
is
io
n
-
s
uppor
t
e
c
os
ys
te
m
th
a
t
e
le
va
te
s
pa
ti
e
nt
c
a
r
e
a
nd a
c
c
e
le
r
a
te
s
bi
om
e
di
c
a
l
di
s
c
ove
r
y.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
A
ut
hor
s
s
ta
te
no f
undi
ng i
nvol
ve
d.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
A
youb Alla
li
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
I
bt
ih
a
l
A
bouc
ha
ba
ka
✓
✓
✓
✓
✓
N
a
ja
t
R
a
f
a
li
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
289
-
299
298
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
A
ut
hor
s
s
ta
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
da
ta
th
a
t
s
uppor
t
th
e
f
in
d
in
gs
of
th
is
s
tu
dy
a
r
e
ope
nl
y
a
va
il
a
bl
e
in
N
E
J
M
R
S
S
F
E
E
D
a
t
ht
tp
s
:/
/ww
w
.ne
jm
.or
g/
r
s
s
-
f
e
e
d/
.
R
E
F
E
R
E
N
C
E
S
[
1]
C
.
B
ur
ga
n,
J
.
K
ow
a
l
s
ki
,
a
nd
W
.
L
i
a
o,
“
D
e
ve
l
opi
ng
a
r
e
t
r
i
e
va
l
a
ugm
e
nt
e
d
ge
ne
r
a
t
i
on
(
R
A
G
)
c
ha
t
bot
a
pp
u
s
i
ng
a
da
pt
i
ve
l
a
r
g
e
l
a
ngua
ge
m
ode
l
s
(
L
L
M
)
a
nd
L
a
ngC
h
a
i
n
f
r
a
m
e
w
or
k,”
P
r
oc
e
e
di
ngs
of
t
he
W
e
s
t
V
i
r
gi
ni
a
A
c
ade
m
y
of
Sc
i
e
nc
e
,
vol
.
96,
no.
1,
2024,
doi
:
10.55632/
pw
va
s
.v96i
1.1068.
[
2]
Y
.
M
a
o,
X
.
D
ong,
W
.
X
u,
Y
.
G
a
o,
B
.
W
e
i
,
a
nd
Y
.
Z
ha
ng,
“
F
I
T
-
R
A
G
:
bl
a
c
k
-
box
R
A
G
w
i
t
h
f
a
c
t
ua
l
i
n
f
or
m
a
t
i
on
a
nd
t
oke
n
r
e
duc
t
i
on,”
A
C
M
T
r
ans
ac
t
i
ons
on I
nf
or
m
at
i
on Sy
s
t
e
m
s
, vol
. 43, no. 2, pp. 1
–
27
, M
a
r
. 2025, doi
:
10.1145/
3676957.
[
3]
D.
-
I
.
M
.
S
c
hi
e
l
e
,
Y
.
G
i
t
t
m
a
nn,
S
.
I
l
c
hm
a
nn,
A
.
G
oj
s
a
l
i
ć
,
D
.
J
u
r
i
nč
i
ć
,
a
nd
P
.
K
l
e
m
pt
,
“
V
ot
i
ng
a
dvi
c
e
a
ppl
i
c
a
t
i
ons
:
i
m
pl
e
m
e
nt
a
t
i
on
of
R
A
G
-
s
uppor
t
e
d L
L
M
s
,”
T
e
c
hR
x
i
v
, J
ul
. 2024, doi
:
10.36227/
t
e
c
hr
xi
v.172115156.64500701/
v1.
[
4]
K
. S
i
ngha
l
e
t
al
.
,
“
T
ow
a
r
d e
xpe
r
t
-
l
e
ve
l
m
e
di
c
a
l
que
s
t
i
on a
ns
w
e
r
i
ng w
i
t
h l
a
r
ge
l
a
ngua
ge
m
ode
l
s
,”
N
at
ur
e
M
e
di
c
i
ne
, vol
. 31,
no. 3
,
pp. 943
–
950, M
a
r
. 2025, doi
:
10.1038/
s
41591
-
024
-
03423
-
7.
[
5]
A
.
Y
.
A
l
a
n,
E
.
K
a
r
a
a
r
s
l
a
n,
a
nd
Ö
.
A
ydı
n,
“
I
m
pr
ov
i
ng
L
L
M
r
e
l
i
a
bi
l
i
t
y
w
i
t
h
R
A
G
i
n
r
e
l
i
gi
ous
que
s
t
i
on
-
a
ns
w
e
r
i
ng:
M
uf
a
s
s
i
r
Q
A
S
,”
T
ur
k
i
s
h J
our
nal
of
E
ngi
ne
e
r
i
ng
, vol
. 9, no. 3, pp. 544
–
559, J
ul
. 2025, doi
:
10.3
1127/
t
uj
e
.1624773.
[
6]
Z
.
X
i
a
o,
X
.
H
e
,
H
.
W
u,
B
.
Y
u,
a
nd
Y
.
G
uo,
“
E
D
A
-
C
opi
l
ot
:
a
R
A
G
-
po
w
e
r
e
d
i
nt
e
l
l
i
ge
nt
a
s
s
i
s
t
a
nt
f
or
E
D
A
t
ool
s
,”
A
C
M
T
r
ans
ac
t
i
ons
on D
e
s
i
gn A
ut
om
at
i
on of
E
l
e
c
t
r
oni
c
Sy
s
t
e
m
s
, vol
. 30, no. 6, pp. 1
–
24, N
ov. 2025, doi
:
10.1145/
3715326.
[
7]
K
.
S
om
a
n
e
t
al
.
,
“
B
i
om
e
di
c
a
l
know
l
e
dge
gr
a
ph
-
opt
i
m
i
z
e
d
p
r
om
pt
ge
ne
r
a
t
i
on
f
or
l
a
r
ge
l
a
ngua
ge
m
ode
l
s
,”
B
i
o
i
nf
or
m
at
i
c
s
,
vol
.
40,
no. 9, S
e
p. 2024, doi
:
10.1093/
bi
oi
nf
or
m
a
t
i
c
s
/
bt
a
e
560.
[
8]
D
.
C
.
-
N
i
e
ve
s
a
nd
L
.
G
.
-
F
or
t
e
,
“
H
um
a
n
-
c
e
nt
e
r
e
d
A
I
f
o
r
m
i
gr
a
nt
i
nt
e
gr
a
t
i
on
t
hr
ough
L
L
M
a
nd
R
A
G
opt
i
m
i
z
a
t
i
on,”
A
ppl
i
e
d
Sc
i
e
nc
e
s
, vol
. 15, no. 1, D
e
c
. 2024, doi
:
10.3390/
a
pp15010325.
[
9]
A
.
M
a
ns
ur
ova
,
A
.
M
a
ns
ur
ova
,
a
nd
A
.
N
ugum
a
nova
,
“
Q
A
-
R
A
G
:
e
xpl
or
i
ng
L
L
M
r
e
l
i
a
nc
e
on
e
xt
e
r
na
l
know
l
e
dge
,”
B
i
g
D
at
a
and
C
ogni
t
i
v
e
C
om
put
i
ng
, vol
. 8, no. 9, S
e
p. 2024, doi
:
10.3390/
bdc
c
8090115.
[
10]
X
.
Z
ha
o,
X
.
Z
hou,
a
nd
G
.
L
i
,
“
C
ha
t
2D
a
t
a
:
a
n
i
nt
e
r
a
c
t
i
ve
d
a
t
a
a
na
l
y
s
i
s
s
ys
t
e
m
w
i
t
h
R
A
G
,
ve
c
t
or
da
t
a
b
a
s
e
s
a
nd
L
L
M
s
,
”
P
r
oc
e
e
di
ngs
of
t
he
V
L
D
B
E
ndow
m
e
nt
, vol
. 17, no. 12, pp. 4481
–
4484, A
ug. 2024, doi
:
10.14778/
3685800.3685905.
[
11]
J
. S
. J
a
uhi
a
i
ne
n a
nd A
.
G
. G
u
e
r
r
a
, “
E
va
l
ua
t
i
ng s
t
ude
nt
s
’
ope
n
-
e
nde
d
w
r
i
t
t
e
n r
e
s
pons
e
s
w
i
t
h
L
L
M
s
:
us
i
ng t
h
e
R
A
G
f
r
a
m
e
w
or
k f
or
G
P
T
-
3.5,
G
P
T
-
4,
C
l
a
ude
-
3,
a
nd
M
i
s
t
r
a
l
-
L
a
r
ge
,”
A
dv
anc
e
s
i
n
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
and
M
ac
hi
ne
L
e
ar
ni
ng
,
vol
.
4,
no.
4,
pp. 3097
–
3113, 2024, doi
:
10.54364/
A
A
I
M
L
.2024.44177.
[
12]
K
.
F
a
ng,
C
.
T
a
ng,
a
nd
J
.
W
a
ng,
“
E
va
l
ua
t
i
ng
s
i
m
ul
a
t
e
d
t
e
a
c
hi
ng
a
udi
o
f
or
t
e
a
c
he
r
t
r
a
i
ne
e
s
us
i
ng
R
A
G
a
nd
l
oc
a
l
L
L
M
s
,”
Sc
i
e
nt
i
f
i
c
R
e
por
t
s
, vol
. 15, no. 1, J
a
n. 2025, doi
:
10.1038/
s
41598
-
025
-
87898
-
5.
[
13]
S
.
V
i
di
ve
l
l
i
,
M
.
R
a
m
a
c
ha
ndr
a
n,
a
nd
A
.
D
ha
r
unba
l
a
j
i
,
“
E
f
f
i
c
i
e
nc
y
-
dr
i
ve
n
c
us
t
om
c
ha
t
bot
de
ve
l
opm
e
nt
:
unl
e
a
s
hi
ng
L
a
ng
C
ha
i
n,
R
A
G
,
a
nd
pe
r
f
or
m
a
nc
e
-
opt
i
m
i
z
e
d
L
L
M
f
us
i
on,
”
C
om
put
e
r
s
,
M
at
e
r
i
al
s
&
C
ont
i
nua
,
vol
.
80,
no.
2,
pp.
2423
–
2442,
2024,
doi
:
10.32604/
c
m
c
.2024.054360.
[
14]
Y
.
W
a
ng,
S
.
L
e
ut
ne
r
,
M
.
I
ngr
i
s
c
h,
C
.
K
l
e
i
n,
L
.
C
.
H
i
n
s
ke
,
a
nd
K
.
D
a
nha
u
s
e
r
,
“
O
pt
i
m
i
z
i
ng
da
t
a
e
xt
r
a
c
t
i
on:
ha
r
ne
s
s
i
ng
R
A
G
a
nd
L
L
M
s
f
or
G
e
r
m
a
n
m
e
di
c
a
l
doc
um
e
nt
s
,”
D
i
gi
t
al
H
e
al
t
h
and
I
nf
or
m
at
i
c
s
I
n
nov
at
i
ons
f
or
Sus
t
ai
nabl
e
H
e
al
t
h
C
ar
e
Sy
s
t
e
m
s
,
vol
. 316, pp. 949
–
950, A
ug. 2024, doi
:
10.3233/
S
H
T
I
240567.
[
15]
R
.
S
.
M
.
W
a
hi
dur
,
S
.
K
i
m
,
H
.
C
hoi
,
D
.
S
.
B
ha
t
t
i
,
a
nd
H
.
-
N
.
L
e
e
,
“
L
e
ga
l
qu
e
r
y
R
A
G
,”
I
E
E
E
A
c
c
e
s
s
,
vol
.
13,
pp.
36978
–
36994
,
2025, doi
:
10.1109/
A
C
C
E
S
S
.2025.3542125.
[
16]
A
.
A
l
l
a
l
i
,
N
.
B
oua
na
ni
,
I
.
A
bouc
ha
ba
ka
,
a
nd
N
.
R
a
f
a
l
i
a
,
“
A
dva
nc
i
ng
e
l
de
r
l
y
c
a
r
e
t
hr
ough
bi
g
da
t
a
a
na
l
yt
i
c
s
a
nd
m
a
c
hi
ne
l
e
a
r
ni
ng
f
or
da
i
l
y
a
c
t
i
vi
t
y
c
ha
r
a
c
t
e
r
i
z
a
t
i
on,”
I
ndone
s
i
an
J
our
nal
of
E
l
e
c
t
r
i
c
al
E
ng
i
ne
e
r
i
ng
and
C
om
put
e
r
Sc
i
e
nc
e
,
vol
.
36,
no.
3,
pp. 1969
–
1975, D
e
c
. 2024, doi
:
10.11591/
i
j
e
e
c
s
.v36.i
3.pp1969
-
1975.
[
17]
M
.
S
on,
Y
.
-
J
.
W
on,
a
nd
S
.
L
e
e
,
“
O
pt
i
m
i
z
i
ng
l
a
r
ge
l
a
ngua
g
e
m
ode
l
s
:
a
de
e
p
d
i
ve
i
nt
o
e
f
f
e
c
t
i
ve
pr
om
pt
e
ngi
ne
e
r
i
ng
t
e
c
hni
que
s
,
”
A
ppl
i
e
d Sc
i
e
nc
e
s
, vol
. 15, no. 3, J
a
n. 2025, doi
:
10.3390/
a
pp15031430.
[
18]
K
.
E
.
K
a
nna
m
m
a
l
,
M
.
R
.
K
.
A
ni
r
udh
,
K
.
P
.
T
a
m
i
z
hi
ni
ya
l
,
G
.
G
a
ni
s
hka
r
,
a
nd
C
.
A
dr
i
na
t
h
,
“
F
i
n
-
R
a
g
a
R
a
g
s
ys
t
e
m
f
or
f
i
na
nc
i
a
l
doc
um
e
nt
s
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
I
nnov
at
i
v
e
Sc
i
e
nc
e
and
R
e
s
e
ar
c
h
T
e
c
h
nol
ogy
,
vol
.
10,
no.
4,
pp.
1761
–
1767,
A
p
r
.
2025,
doi
:
10.38124/
i
j
i
s
r
t
/
25a
pr
1147.
[
19]
P
.
P
a
ny,
“
R
e
a
s
oni
ng
e
ngi
ne
w
i
t
h
pr
e
-
t
r
a
i
ne
d
L
L
M
s
:
a
n
ope
r
a
t
i
on
G
P
T
,”
I
nt
e
r
nat
i
onal
J
our
nal
f
or
R
e
s
e
ar
c
h
i
n
A
ppl
i
e
d
Sc
i
e
nc
e
and E
ngi
ne
e
r
i
ng T
e
c
hnol
ogy
, vol
. 13, no. 4, pp. 2452
–
2463, A
pr
. 2025, doi
:
10.22214/
i
j
r
a
s
e
t
.2025.68761.
[
20]
J
.
W
a
ng
e
t
al
.
,
“
H
i
e
r
a
r
c
hi
c
a
l
i
nde
x
r
e
t
r
i
e
va
l
-
dr
i
ve
n
w
i
r
e
l
e
s
s
ne
t
w
or
k
i
nt
e
nt
t
r
a
ns
l
a
t
i
on
w
i
t
h
L
L
M
,”
I
E
E
E
T
r
ans
ac
t
i
ons
on
M
obi
l
e
C
om
put
i
ng
, vol
. 24, no. 10, pp. 9837
–
9851, O
c
t
. 2025, doi
:
10.1109/
T
M
C
.2025.3564937.
[
21]
A
.
S
ghi
r
,
A
.
A
l
l
a
l
i
,
N
.
R
a
f
a
l
i
a
,
a
nd
J
.
A
bouc
ha
ba
ka
,
“
A
dva
nc
e
d
s
t
r
a
t
e
gi
e
s
f
or
bi
g
da
t
a
r
e
s
our
c
e
a
nd
s
t
or
a
ge
opt
i
m
i
z
a
t
i
on:
a
n
A
I
pe
r
s
pe
c
t
i
ve
,”
I
nt
e
r
nat
i
onal
J
ou
r
nal
of
A
dv
anc
e
d
C
om
put
e
r
Sc
i
e
nc
e
and
A
ppl
i
c
at
i
ons
,
vol
.
16,
no.
8,
2025,
doi
:
10.14569/
I
J
A
C
S
A
.2025.0160896.
[
22]
C
.
C
a
r
pe
nt
e
r
,
“
Z
e
r
o
-
s
hot
l
e
a
r
ni
ng
w
i
t
h
l
a
r
ge
l
a
ngua
ge
m
ode
l
s
e
nha
nc
e
s
dr
i
l
l
i
ng
-
i
nf
or
m
a
t
i
on
r
e
t
r
i
e
va
l
,”
J
our
nal
of
P
e
t
r
ol
e
um
T
e
c
hnol
ogy
, vol
. 77, no. 1, pp. 92
–
95, J
a
n. 2025, doi
:
10.2118/
0125
-
0092
-
J
P
T
.
[
23]
A
.
A
l
l
a
l
i
,
Z
.
E
.
F
a
l
a
h,
A
.
S
ghi
r
,
J
.
A
bouc
ha
ba
ka
,
a
nd
N
.
R
a
f
a
l
i
a
,
“
A
c
om
pa
r
a
t
i
ve
a
na
l
ys
i
s
of
G
P
U
s
,
T
P
U
s
,
D
P
U
s
,
a
nd
Q
P
U
s
f
o
r
de
e
p
l
e
a
r
ni
ng
w
i
t
h
pyt
hon,”
I
ndone
s
i
an
J
our
nal
of
E
l
e
c
t
r
i
c
al
E
ngi
ne
e
r
i
ng
and
C
om
put
e
r
Sc
i
e
nc
e
,
vol
.
38,
no.
2,
pp.
1324
–
1330,
M
a
y 2025, doi
:
10.11591/
i
j
e
e
c
s
.v38.i
2.pp1324
-
1330.
[
24]
W
.
B
i
e
t
al
.
,
“
L
e
ve
r
a
gi
ng
t
he
dua
l
c
a
pa
bi
l
i
t
i
e
s
of
L
L
M
:
L
L
M
-
e
nh
a
nc
e
d
t
e
xt
m
a
ppi
ng
m
ode
l
f
or
pe
r
s
ona
l
i
t
y
de
t
e
c
t
i
on,”
P
r
oc
e
e
di
ngs
of
t
he
A
A
A
I
C
onf
e
r
e
nc
e
on
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
,
vol
.
39,
no.
22,
pp.
23487
–
23495,
A
pr
.
2025,
doi
:
10.1609/
a
a
a
i
.v39i
22.34517.
[
25]
V
.
M
a
l
i
k,
“
H
a
doop
di
s
t
r
i
but
e
d
f
i
l
e
s
ys
t
e
m
(
H
D
F
S
)
w
i
t
h
i
t
s
a
r
c
hi
t
e
c
t
ur
e
,”
I
nt
e
r
nat
i
onal
J
our
nal
f
or
R
e
s
e
ar
c
h
i
n
A
ppl
i
e
d
Sc
i
e
nc
e
and E
ngi
ne
e
r
i
ng T
e
c
hnol
ogy
, vol
. 13, no. 5, pp. 6031
–
6034, M
a
y 2025, doi
:
10
.22214/
i
j
r
a
s
e
t
.2025.71584.
[
26]
S
.
A
w
a
s
t
hi
a
nd
N
.
K
ohl
i
,
“
H
ybr
i
d
e
nc
r
ypt
i
on
f
or
f
or
t
i
f
yi
ng
H
D
F
S
da
t
a
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
B
as
i
c
and
A
ppl
i
e
d
Sc
i
e
nc
e
s
,
vol
. 14, no. 5, pp. 436
–
454, S
e
p. 2025, doi
:
10.14419/
m
46f
n971.
Evaluation Warning : The document was created with Spire.PDF for Python.