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
. 14, No. 5, O
c
to
be
r
2025
, pp.
4202
~
4210
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
5
.pp
4202
-
4210
4202
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
L
e
ar
n
i
n
g ass
i
st
an
c
e
m
od
u
l
e
b
ase
d
o
n
a sm
al
l
l
a
n
gu
age
m
od
e
l
M
ar
c
o A
n
t
on
io
Ji
n
e
t
e
1
, R
ob
in
s
on
Ji
m
é
n
e
z
-
M
or
e
n
o
1
, A
n
n
y
A
s
t
r
id
E
s
p
it
ia
-
C
u
b
il
lo
s
2
1
M
e
c
ha
t
r
oni
c
E
ngi
ne
e
r
i
ng P
r
ogr
a
m
,
E
ngi
ne
e
r
i
ng
F
a
c
ul
t
y
,
U
ni
ve
r
s
i
da
d M
i
l
i
t
a
r
N
ue
va
G
r
a
na
da
,
B
ogot
á
,
C
ol
om
bi
a
2
I
ndus
t
r
i
a
l
E
ngi
ne
e
r
i
ng P
r
ogr
a
m
,
E
ngi
ne
e
r
i
ng
F
a
c
ul
t
y
,
U
ni
ve
r
s
i
da
d
M
i
l
i
t
a
r
N
ue
va
G
r
a
na
da
,
B
ogot
á
,
C
ol
om
bi
a
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
a
n 23, 2025
R
e
vi
s
e
d
J
ul
11, 2025
A
c
c
e
pt
e
d
A
ug 6, 2025
This
paper
presents
the
development
of
a
low
-
cost
learning
as
sistant
embedded
in
an
NVIDIA
Jetson
Xavier
board
that
uses
speech
and
gesture
recognition
,
together
with
a
long
language
model
for
offline
work.
Usi
ng
the
large
language
model
(
LLM
)
Phi
-
3
Mini
(3
.
8B)
model
and
the
Whisper
(
model
base
)
model
for
automatic
speech
recognition,
a
learning
assis
tant
is
obtained
under
a
compact
and
efficient
design
based
on
extensive
la
nguage
model
architectures
that
give
a
general
answer
set
of
a
topic.
Average
processing
times
of
0.108
seconds
per
character,
a
speech
transc
ription
efficiency
of
94.75%,
an
average
accuracy
of
9.5/10
and
8.5/10
in
the
consist
ency
of
the
r
esponses
generated
by
the
learning
assistan
t,
a
full
recognition
of
the
hand
raising
gesture
when
done
for
at
least
2
se
conds,
even
without
fully
extending
the
fingers,
were
obtained.
The
prototype
is
based
on
the
design
of
a
graphical
interface
capable
of
responding
to
voice
commands
and
generating
dynamic
interacti
ons
in
response
to
the
user'
s
ge
sture
detection,
representing
a
significant
advance
towards
the
creation
of
comprehens
ive and acces
sible h
uman
-
machine inte
rface
solutions.
K
e
y
w
o
r
d
s
:
A
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
D
e
e
p l
e
a
r
ni
ng
E
m
be
dde
d s
ys
te
m
L
a
r
ge
l
a
ngua
ge
m
ode
l
L
e
a
r
ni
ng a
s
s
is
ta
nt
S
m
a
ll
la
ngua
ge
m
ode
l
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
nny As
tr
id
E
s
pi
ti
a
-
C
ubi
ll
os
I
ndus
tr
ia
l
E
ngi
ne
e
r
in
g P
r
og
r
a
m
, E
ngi
ne
e
r
in
g F
a
c
ul
ty
,
U
ni
ve
r
s
i
da
d
M
il
it
a
r
N
ue
va
G
r
a
na
da
C
a
r
r
e
r
a
11 #101
-
80, B
ogot
á
, C
ol
om
bi
a
E
m
a
il
:
a
nny
.e
s
pi
ti
a
@
uni
m
il
it
a
r
.e
du.c
o
1.
I
N
T
R
O
D
U
C
T
I
O
N
A
dva
nc
e
s
in
na
tu
r
a
l
la
ngua
ge
m
ode
ls
a
nd
th
e
ir
a
ppl
ic
a
ti
ons
w
it
h
la
r
ge
la
ngua
ge
m
ode
l
(
L
L
M
)
th
a
t
a
ll
ow
pe
opl
e
to
in
te
r
a
c
t
to
da
y
in
a
m
or
e
na
tu
r
a
l
w
a
y
w
it
h
c
om
put
e
r
s
a
nd
r
obot
s
is
boomi
ng
a
nd
in
f
ul
l
de
ve
lo
pm
e
nt
[
1]
.
S
o
th
a
t
to
da
y
it
is
pos
s
ib
le
to
f
in
d
r
e
s
e
a
r
c
h
w
it
h
a
dva
nc
e
s
in
di
f
f
e
r
e
nt
f
ie
ld
s
s
uc
h
a
s
a
s
s
is
t
a
nc
e
s
y
s
te
m
s
f
or
in
dus
tr
ia
l
pr
oc
e
s
s
e
s
[
2]
,
r
obot
c
ont
r
ol
[
3]
,
voi
c
e
a
s
s
is
ta
nt
s
in
ta
s
ks
s
u
c
h
a
s
m
e
di
c
a
l
di
a
gnos
is
[
4]
,
a
nd
de
c
i
s
io
n
m
a
ki
ng
in
m
a
nuf
a
c
tu
r
in
g
pr
oc
e
s
s
e
s
[
5]
,
[
6]
.
T
he
in
te
gr
a
ti
on
of
L
L
M
'
s
w
it
h
ot
he
r
in
f
or
m
a
ti
on
m
a
na
ge
m
e
nt
s
ys
te
m
s
s
u
c
h
a
s
C
ha
tG
P
T
[
7]
a
ll
ow
s
th
e
de
v
e
lo
pm
e
nt
of
e
ve
n
m
or
e
s
pe
c
i
a
li
z
e
d
a
ppl
ic
a
ti
ons
in
a
r
e
a
s
s
uc
h
a
s
m
e
di
c
a
l
or
th
ope
di
c
di
a
gno
s
ti
c
s
[
8]
.
I
n
tu
r
n,
ot
he
r
to
ol
s
a
r
e
us
e
d
s
uc
h
a
s
a
ugm
e
nt
e
d
r
e
a
li
ty
to
s
uppor
t
e
m
e
r
ge
nc
y
r
e
s
pon
s
e
[
9]
,
c
om
put
e
r
vi
s
io
n
s
ys
te
m
s
f
or
opht
ha
lm
ol
ogy
a
s
s
is
ta
nt
s
[
10]
or
th
e
in
te
gr
a
ti
on
of
pr
om
pt
e
ngi
ne
e
r
in
g
te
c
hni
que
s
s
uc
h
a
s
r
e
tr
ie
va
l
-
a
ugm
e
nt
e
d
g
e
ne
r
a
ti
on
(
R
A
G
)
,
a
nd
in
c
or
por
a
ti
ng doma
in
-
s
pe
c
if
ic
knowle
dge
gr
a
phs
(
K
G
s
)
[
11]
.
H
ow
e
ve
r
,
s
pe
c
if
ic
s
c
h
e
m
e
s
of
L
L
M
us
e
a
ll
ow
c
om
pl
e
m
e
nt
in
g
im
por
ta
nt
de
ve
lo
pm
e
nt
s
,
s
uc
h
a
s
di
s
in
f
or
m
a
ti
on
or
f
a
ls
e
in
f
or
m
a
ti
on
f
r
o
m
th
e
in
te
r
ne
t
[
12]
,
c
yb
e
r
th
r
e
a
ts
[
13]
,
or
ge
ne
r
a
ti
ng
c
ouns
e
li
ng
to
ol
s
f
or
pe
opl
e
[
14
]
.
L
L
M
s
a
r
e
be
in
g
us
e
d
a
s
a
s
s
i
s
ta
nc
e
to
ol
s
[
15]
,
in
c
lu
di
ng
c
ha
tb
ot
s
[
16]
.
W
it
h
de
di
c
a
te
d
ha
r
dw
a
r
e
im
pl
e
m
e
nt
a
ti
ons
[
17]
,
th
e
y c
a
n
of
f
e
r
s
pe
c
if
ic
s
ol
ut
io
n
s
in
di
f
f
e
r
e
nt
a
r
e
a
s
of
knowle
dge
. T
hi
s
c
a
n
b
e
c
om
pl
e
m
e
nt
e
d
w
it
h
le
a
r
ni
ng
a
s
s
is
ta
nt
s
s
uc
h
a
s
th
e
one
pr
opos
e
d
in
th
is
w
or
k,
w
hi
c
h
in
te
gr
a
te
s
a
ut
om
a
ti
c
s
pe
e
c
h
r
e
c
ogni
ti
on
a
nd
c
om
put
e
r
vi
s
io
n
s
ys
te
m
s
,
a
s
a
c
om
pl
e
m
e
nt
to
th
e
s
ta
te
of
th
e
a
r
t.
A
ddi
ti
ona
ll
y
a
nd
us
in
g
de
di
c
a
te
d
ha
r
dw
a
r
e
c
a
r
ds
de
ve
lo
pe
d
by
N
V
I
D
I
A
th
a
t
s
uppor
t
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
lg
or
it
hm
s
,
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
L
e
ar
ni
ng as
s
i
s
ta
nc
e
m
odul
e
ba
s
e
d on a s
m
al
l
la
nguage
m
od
e
l
(
M
ar
c
o A
nt
oni
o J
in
e
te
)
4203
th
os
e
pr
e
s
e
nt
e
d
in
[
18]
–
[
20]
,
L
L
M
s
w
it
h
s
m
a
ll
m
ode
l
s
c
a
n
a
ls
o
be
a
da
pt
e
d
[
21]
.
T
he
r
e
f
or
e
,
it
is
pr
opos
e
d
to
de
s
ig
n
a
n
e
m
be
dd
e
d
le
a
r
ni
ng
a
s
s
is
ta
nt
th
a
t
e
na
bl
e
s
a
ut
o
m
a
ti
c
ge
s
tu
r
e
a
nd
s
pe
e
c
h
r
e
c
ogni
ti
on.
T
he
c
ont
r
ib
ut
io
n of
t
hi
s
w
or
k i
s
f
oc
us
e
d on the
i
m
pl
e
m
e
nt
a
ti
on of
a
l
e
a
r
ni
ng modul
e
t
ha
t
doe
s
not
r
e
qui
r
e
i
nt
e
r
ne
t
c
onne
c
ti
on a
nd s
uppor
ts
t
he
ge
ne
r
a
ti
on of
ge
ne
r
a
l
knowle
dge
a
ns
w
e
r
s
unde
r
a
n e
a
s
y t
o us
e
i
nt
e
r
f
a
c
e
, t
hr
ough
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
lg
or
it
hm
s
s
uc
h
a
s
lo
ng
m
ode
l
la
ngua
g
e
a
nd
m
ot
io
n
c
a
pt
ur
e
a
nd
s
p
e
e
c
h
r
e
c
ogni
ti
on
s
ys
te
m
s
f
or
na
tu
r
a
l
in
te
r
a
c
ti
on,
s
o
th
a
t
it
i
s
hypothe
s
iz
e
d
th
a
t
a
n
e
m
be
dde
d
l
e
a
r
ni
ng
s
ys
t
e
m
c
a
n
f
a
c
il
it
a
te
que
r
ie
s
f
r
om
pe
opl
e
f
or
l
e
a
r
ni
ng t
ha
t
do not ha
ve
i
nt
e
r
ne
t
c
onne
c
ti
on, s
uc
h a
s
r
ur
a
l
a
r
e
a
s
.
T
hi
s
p
a
pe
r
is
s
tr
uc
tu
r
e
d
in
f
our
s
e
c
ti
ons
;
th
e
f
ir
s
t
one
pr
e
s
e
nt
s
th
e
s
ta
t
e
of
th
e
a
r
t
a
nd
th
e
pr
opos
e
d
w
or
k. T
he
s
e
c
ond
s
e
c
ti
on de
s
c
r
ib
e
s
t
he
m
e
th
odol
ogy de
ve
lo
pe
d
f
or
t
he
de
s
ig
n of
a
l
ow
-
c
os
t
le
a
r
ni
ng a
s
s
is
t
a
nt
f
or
a
ut
om
a
ti
c
ge
s
tu
r
e
a
nd
s
pe
e
c
h
r
e
c
ogni
ti
on.
T
he
th
ir
d
s
e
c
ti
o
n
pr
e
s
e
nt
s
th
e
a
na
ly
s
i
s
a
nd
di
s
c
u
s
s
io
n
of
th
e
r
e
s
ul
ts
obt
a
in
e
d.
F
in
a
ll
y,
th
e
f
our
th
s
e
c
ti
on
pr
e
s
e
nt
s
th
e
c
on
c
lu
s
io
ns
de
r
iv
e
d
f
r
om
th
e
te
s
t
r
e
s
ul
ts
of
th
e
pr
ot
ot
ype
de
ve
lo
pe
d i
n t
hi
s
r
e
s
e
a
r
c
h.
2.
M
E
T
H
O
D
W
it
h
th
e
pur
pos
e
of
de
ve
lo
pi
ng
a
pr
ot
ot
ype
te
a
c
hi
ng
a
s
s
is
ta
nt
th
r
ough
th
e
m
os
t
na
tu
r
a
l
in
te
r
a
c
ti
on
f
e
a
s
ib
le
a
t
lo
w
c
os
t,
f
iv
e
pha
s
e
s
w
e
r
e
e
s
ta
bl
i
s
he
d.
T
he
f
ir
s
t
to
de
f
in
e
th
e
m
os
t
c
onve
ni
e
nt
to
ol
f
or
s
pe
e
c
h
tr
a
ns
c
r
ip
ti
on, t
he
s
e
c
ond to s
e
le
c
t
th
e
t
ool
f
or
r
e
s
pons
e
ge
n
e
r
a
ti
on, t
he
t
hi
r
d f
or
ge
s
tu
r
e
r
e
c
ogni
ti
on, t
he
f
our
th
to
in
te
gr
a
te
a
ll
th
e
to
ol
s
th
r
ough
a
n
opt
im
iz
e
d
gr
a
phi
c
a
l
in
te
r
f
a
c
e
,
a
nd
f
in
a
ll
y
th
e
f
if
th
pha
s
e
to
va
li
da
te
th
e
ope
r
a
ti
on
of
th
e
pr
ot
ot
ype
us
in
g
lo
w
e
ne
r
gy
c
ons
um
pt
io
n
d
e
vi
c
e
s
.
T
e
n
te
s
t
s
c
e
na
r
io
s
a
r
e
e
s
t
a
bl
is
he
d
to
va
li
da
te
th
e
r
e
s
ul
ts
of
th
e
voi
c
e
in
te
r
a
c
ti
on
m
ode
l
a
nd
10
s
c
e
na
r
io
s
to
e
va
lu
a
t
e
th
e
r
e
s
pons
e
s
of
th
e
la
ngu
a
ge
m
ode
l
us
e
d,
obt
a
in
in
g
m
e
tr
ic
s
s
uc
h
a
s
a
ve
r
a
ge
r
e
s
pon
s
e
ti
m
e
,
a
c
c
ur
a
c
y
,
a
nd
c
on
s
is
te
nc
y.
F
ig
ur
e
1
s
how
s
th
e
f
lo
w
di
a
gr
a
m
of
t
he
m
e
th
odol
ogy pr
opos
e
d f
or
t
hi
s
r
e
s
e
a
r
c
h.
F
ig
ur
e
1. M
e
th
odol
ogy f
lo
w
c
ha
r
t
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
3.1.
S
p
e
e
c
h
t
r
an
s
c
r
ip
t
io
n
A
m
ong
di
f
f
e
r
e
nt
te
xt
t
r
a
ns
c
r
ip
ti
on
to
ol
s
s
uc
h
a
s
B
e
a
r
F
il
e
C
onve
r
te
r
,
D
ic
ta
ti
on
a
nd
G
oogl
e
'
s
G
boa
r
d,
W
hi
s
pe
r
de
ve
lo
pe
d
by
O
p
e
nA
I
s
ta
nd
s
out
f
or
it
s
e
f
f
ic
i
e
nc
y,
por
ta
bi
li
ty
a
nd
f
r
e
e
dom
of
us
e
due
to
it
s
ope
n
-
s
our
c
e
c
ode
.
I
t
is
tr
a
in
e
d
w
it
h
m
or
e
th
a
n
one
m
il
li
on
hour
s
of
a
udi
o
in
it
s
th
ir
d
ve
r
s
io
n
a
nd a
n
e
r
r
or
r
a
te
of
le
s
s
th
a
n
5%
in
S
pa
ni
s
h
-
la
ngua
ge
tr
a
ns
c
r
ip
ti
ons
,
c
ons
id
e
r
in
g
punc
tu
a
ti
on
m
a
r
ks
s
uc
h
a
s
c
om
m
a
s
a
nd
pe
r
io
ds
.
I
ts
ba
s
ic
s
tr
uc
tu
r
e
,
a
s
s
how
n
in
F
ig
ur
e
2,
ha
s
tr
a
ns
f
or
m
e
r
e
nc
ode
r
/d
e
c
ode
r
bl
oc
ks
,
ba
s
e
d
on
a
s
pe
c
tr
ogr
a
m
i
nput
f
r
om
t
he
a
udi
o s
our
c
e
[
22]
, [
23]
.
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
. 14,
N
o. 5, Oc
to
be
r
2025
:
4202
-
4210
4204
F
ig
ur
e
2
. R
e
c
ogni
ti
on a
r
c
hi
te
c
tu
r
e
by W
hi
s
pe
r
[
24]
3.2.
L
an
gu
age
m
od
e
l
P
hi
-
3
M
in
i
de
ve
lo
pe
d
by
M
ic
r
os
of
t
s
ta
nds
out
a
s
a
s
m
a
ll
la
n
gua
ge
m
ode
l
(
S
M
L
)
la
ngua
ge
m
ode
l
c
om
pa
r
a
bl
e
to
lo
ng
la
ngu
a
ge
m
ode
l
s
s
uc
h
a
s
C
ha
tG
P
T
[
25]
.
W
hi
c
h
is
u
s
e
d
in
it
s
3.8
B
pa
r
a
m
e
te
r
la
ngua
ge
m
ode
l
ve
r
s
io
n,
a
va
il
a
bl
e
in
two
c
ont
e
xt
le
ngt
h
s
of
128
k
a
nd
4
k
to
ke
ns
or
a
to
m
ic
pa
r
ts
of
th
e
la
ngua
ge
th
a
t
pr
oc
e
s
s
e
s
th
e
m
ode
l,
a
ll
ow
in
g
a
qu
e
s
ti
on
to
b
e
a
s
ke
d
to
th
e
m
ode
l,
w
it
h
m
or
e
r
e
le
va
nt
a
ns
w
e
r
s
f
r
om
th
e
m
ode
l
a
nd
a
w
id
e
va
r
ie
ty
of
c
ont
e
nt
g
e
ne
r
a
ti
on
c
ont
e
xt
(
s
e
e
F
ig
ur
e
3)
.
T
hi
s
m
ode
l
is
c
ho
s
e
n
f
or
c
ont
e
nt
ge
ne
r
a
ti
on
be
c
a
us
e
of
it
s
s
m
a
ll
s
to
r
a
ge
s
iz
e
a
nd
th
e
a
s
s
oc
ia
t
e
d
de
ve
lo
pm
e
nt
s
w
it
h
N
V
I
D
I
A
ha
r
dw
a
r
e
to
obt
a
in
l
oc
a
l
m
ode
ls
.
F
ig
ur
e
3. C
ont
e
nt
ge
ne
r
a
ti
on by the
P
hi
-
3 M
in
i
m
ode
l
[
25]
D
e
s
pi
te
it
s
s
m
a
ll
e
r
s
iz
e
c
om
pa
r
e
d
to
la
r
ge
r
m
ode
ls
,
P
hi
-
3
M
i
ni
s
ta
nds
out
f
or
it
s
ba
la
nc
e
be
twe
e
n
pe
r
f
or
m
a
nc
e
,
s
pe
e
d
a
nd
c
om
put
a
ti
ona
l
r
e
qui
r
e
m
e
nt
s
,
m
a
ki
n
g
it
a
n
a
f
f
or
da
bl
e
a
nd
ve
r
s
a
ti
le
s
ol
ut
io
n
f
or
di
ve
r
s
e
ne
e
ds
.
T
he
te
xt
-
to
-
s
pe
e
c
h
(
T
T
S
)
m
odul
e
de
ve
lo
pe
d
in
th
is
r
e
s
e
a
r
c
h
in
te
gr
a
te
s
th
e
G
oogl
e
te
xt
-
to
-
s
pe
e
c
h
(
gT
T
S
)
a
nd
pl
a
ys
ound
li
br
a
r
ie
s
to
pr
ovi
de
a
n
e
f
f
ic
ie
nt
s
ol
ut
io
n
f
or
s
ynt
he
s
iz
in
g
a
nd
r
e
pr
oduc
in
g
r
e
s
pons
e
s
ge
n
e
r
a
te
d
by
la
ngua
ge
m
ode
l
s
.
I
ts
im
p
le
m
e
nt
a
ti
on
is
de
s
ig
ne
d
to
opt
im
iz
e
r
e
a
l
-
ti
m
e
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
L
e
ar
ni
ng as
s
i
s
ta
nc
e
m
odul
e
ba
s
e
d on a s
m
al
l
la
nguage
m
od
e
l
(
M
ar
c
o A
nt
oni
o J
in
e
te
)
4205
in
te
r
a
c
ti
on
be
twe
e
n
th
e
us
e
r
a
nd
th
e
s
ys
te
m
, c
ons
ol
id
a
ti
ng
it
s
f
unc
ti
ona
li
ty
w
it
hi
n
th
e
N
V
I
D
I
A
J
e
t
s
on
X
a
vi
e
r
A
G
X
e
nvi
r
onm
e
nt
.
T
o
r
e
duc
e
pe
r
c
e
iv
e
d
la
t
e
nc
y
a
nd
im
pr
ove
us
e
r
e
xpe
r
ie
nc
e
,
a
t
e
xt
f
r
a
gm
e
nt
a
ti
on
s
tr
a
te
gy,
known
a
s
c
hunking,
w
a
s
im
pl
e
m
e
nt
e
d.
T
hi
s
s
tr
a
te
gy
di
vi
de
s
th
e
c
ont
e
nt
in
to
m
a
na
ge
a
bl
e
bl
oc
ks
of
a
ppr
oxi
m
a
te
ly
80
to
100
c
ha
r
a
c
te
r
s
.
T
hi
s
a
ppr
oa
c
h
e
n
s
ur
e
s
th
a
t
e
a
c
h
bl
oc
k
i
s
pr
oc
e
s
s
e
d
a
nd
pl
a
ye
d
s
e
que
nt
ia
ll
y,
a
ll
ow
in
g
a
udi
o
out
put
to
be
gi
n
qui
c
kl
y
w
it
hout
th
e
ne
e
d
to
pr
oc
e
s
s
th
e
e
nt
ir
e
te
xt
be
f
or
e
ha
nd.
T
he
t
e
xt
is
di
vi
de
d
in
to
w
or
ds
a
nd
dyna
m
ic
a
ll
y
gr
oupe
d
unt
il
th
e
de
f
in
e
d
li
m
it
is
r
e
a
c
he
d,
e
n
s
ur
in
g
th
a
t
e
a
c
h
bl
oc
k
c
ont
a
in
s
a
ba
la
nc
e
d
num
be
r
of
w
or
ds
.
O
nc
e
gr
oupe
d,
th
e
bl
oc
k
is
pr
oc
e
s
s
e
d
by
a
f
unc
ti
on
th
a
t
ge
ne
r
a
te
s
th
e
a
udi
o
a
nd
pl
a
ys
i
t
im
m
e
di
a
te
ly
, pr
ovi
di
ng a
s
e
a
m
le
s
s
a
nd c
ont
in
uou
s
e
xpe
r
ie
nc
e
.
E
a
c
h
te
xt
bl
oc
k
is
c
onve
r
te
d
to
a
udi
o
us
in
g
gT
T
S
,
c
onf
ig
ur
e
d
to
ge
ne
r
a
te
S
pa
ni
s
h
voi
c
e
(
la
ng=
'
e
s
'
)
a
nd
a
dj
us
te
d
to
a
s
ta
nd
a
r
d
a
c
c
e
nt
th
r
ough
th
e
tl
d=
'
us
'
pa
r
a
m
e
t
e
r
.
T
he
r
e
s
ul
ti
ng
a
udi
o
i
s
te
m
por
a
r
il
y
s
to
r
e
d
in
M
P
3
f
or
m
a
t
us
in
g
th
e
te
m
pf
i
le
li
br
a
r
y,
e
ns
ur
in
g
c
om
pa
ti
bi
li
t
y
w
it
h
th
e
pl
a
ys
ound
pl
a
yba
c
k
s
ys
te
m
.
T
he
pr
oc
e
s
s
in
g
f
unc
ti
on
e
ns
ur
e
s
th
a
t
th
e
t
e
m
por
a
r
y
f
il
e
is
pl
a
y
e
d
im
m
e
di
a
te
ly
a
f
te
r
ge
n
e
r
a
ti
on,
opt
im
iz
in
g
r
e
s
our
c
e
us
a
ge
a
nd gua
r
a
nt
e
e
in
g unint
e
r
r
upt
e
d us
e
r
i
nt
e
r
a
c
ti
on.
3.3.
G
e
s
t
u
r
e
r
e
c
ogn
it
io
n
(
r
ai
s
e
d
h
an
d
)
G
e
s
tu
r
e
r
e
c
ogni
ti
on,
s
pe
c
if
ic
a
ll
y
th
e
de
te
c
ti
on
of
a
r
a
is
e
d
ha
nd
,
w
a
s
im
pl
e
m
e
nt
e
d
us
in
g
M
e
di
a
P
ip
e
ne
ur
a
l
ne
twor
k
-
ba
s
e
d
s
ol
ut
io
n
de
s
ig
ne
d
to
tr
a
c
k
ke
y
poi
nt
s
on
th
e
ha
nd
[
26]
.
T
hi
s
a
ppr
oa
c
h
a
ll
ow
s
a
c
c
ur
a
te
id
e
nt
if
ic
a
ti
on
of
ha
nd
pos
it
io
n a
nd
m
ove
m
e
nt
in
r
e
a
l
ti
m
e
,
pr
ovi
di
ng
a
s
ol
id
b
a
s
is
f
or
in
te
r
a
c
ti
ng
dyna
m
ic
a
ll
y
w
it
h
th
e
s
ys
te
m
.
T
he
c
onf
ig
ur
a
ti
on
in
c
lu
de
s
a
2
-
s
e
c
ond
ti
m
e
th
r
e
s
hol
d,
w
hi
c
h
a
c
ts
a
s
a
c
r
it
e
r
io
n
f
o
r
c
onf
ir
m
in
g
th
e
ge
s
tu
r
e
a
nd
a
c
ti
va
ti
ng
th
e
c
or
r
e
s
ponding
f
unc
ti
ons
.
T
hi
s
m
e
th
odol
ogy
e
ns
ur
e
s
th
a
t
in
te
r
a
c
ti
ons
a
r
e
in
te
nt
io
na
l
a
nd
a
voi
d
a
c
c
id
e
nt
a
l
a
c
ti
va
ti
ons
.
T
he
r
e
c
ogni
ti
on
is
r
obus
t
s
in
c
e
it
w
or
ks
e
ve
n
w
it
hout
s
how
in
g a
ll
t
he
f
in
ge
r
s
of
t
he
ha
nd e
xt
e
nde
d.
3.4.
O
p
t
im
iz
e
d
gr
ap
h
ic
i
n
t
e
r
f
ac
e
T
hi
s
s
e
c
ti
on
d
e
ta
il
s
th
e
de
s
ig
n
of
th
e
in
te
r
f
a
c
e
,
it
s
t
e
c
hni
c
a
l
c
ha
r
a
c
te
r
is
ti
c
s
,
a
nd
th
e
im
pl
e
m
e
nt
e
d
w
or
kf
lo
w
,
in
c
lu
di
ng
c
a
m
e
r
a
in
te
r
a
c
ti
on
th
r
ough
th
e
r
e
c
ogni
ti
on
of
r
a
is
e
d
ha
nds
.
I
t
a
ls
o
di
s
c
us
s
e
s
how
th
e
s
ys
te
m
w
a
s
opt
im
iz
e
d
to
e
ns
ur
e
lo
w
r
e
s
our
c
e
c
ons
um
pt
io
n,
pr
io
r
it
iz
in
g
f
unc
ti
ona
li
ty
a
nd
us
e
r
e
xp
e
r
ie
nc
e
in
th
e
c
ont
e
xt
of
c
om
pa
c
t
a
nd e
f
f
ic
ie
nt
ha
r
dw
a
r
e
. A
c
c
or
di
ngl
y,
us
in
g a
n
N
V
I
D
I
A
J
e
ts
on X
a
vi
e
r
A
G
X
c
a
r
d
[
27]
,
th
e
W
hi
s
pe
r
m
ode
l
f
or
s
pe
e
c
h
r
e
c
ogni
ti
on
a
nd
P
hi
-
3
M
in
i
a
s
a
na
tu
r
a
l
la
ngua
ge
m
ode
l
a
r
e
im
pl
e
m
e
nt
e
d.
T
he
N
V
I
D
I
A
J
e
t
s
on X
a
vi
e
r
A
G
X
c
a
r
d us
e
d ha
s
t
he
f
ol
lo
w
in
g s
pe
c
if
ic
a
ti
ons
:
–
C
P
U
:
8
-
c
or
e
A
R
M
v8.2
–
G
P
U
:
512
-
c
o
r
e
vol
ta
G
P
U
w
it
h
te
ns
or
c
or
e
s
–
M
e
m
or
y:
16
G
B
L
P
D
D
R
4x
–
S
to
r
a
ge
:
32
G
B
e
M
M
C
5.1
A
s
im
pl
e
gr
a
phi
c
a
l
in
te
r
f
a
c
e
w
a
s
de
ve
lo
pe
d,
pr
e
s
e
nt
e
d
in
F
ig
ur
e
4,
w
it
h
two
t
e
xt
boxe
s
:
one
f
or
voi
c
e
tr
a
ns
c
r
ip
ti
on
(
W
hi
s
pe
r
)
a
nd
one
f
or
L
L
M
r
e
s
pons
e
(
P
hi
-
3
M
in
i)
;
a
nd
it
in
c
lu
de
s
a
s
ta
r
t
but
to
n
th
a
t
a
c
ti
va
te
s
t
he
e
nt
ir
e
pr
oc
e
s
s
(
ge
s
tu
r
e
de
t
e
c
ti
on, S
T
T
,
a
nd L
L
M
)
. T
he
s
y
s
te
m
c
om
bi
ne
s
g
e
s
tu
r
e
de
te
c
ti
on, voic
e
tr
a
ns
c
r
ip
ti
on
a
nd
s
pe
e
c
h
ge
ne
r
a
ti
on
c
a
p
a
bi
li
ti
e
s
in
to
one
in
te
gr
a
te
d
f
lo
w
.
T
he
pr
ogr
a
m
is
de
s
ig
ne
d
to
s
t
a
r
t
a
ut
om
a
ti
c
a
ll
y
w
he
n
th
e
N
V
I
D
I
A
J
e
ts
on
X
a
vi
e
r
A
G
X
is
tu
r
ne
d
on,
e
ns
ur
in
g
th
a
t
th
e
s
ys
te
m
is
r
e
a
dy
f
or
us
e
w
it
hout
r
e
qui
r
in
g a
ddi
ti
ona
l
c
onf
ig
ur
a
ti
on.
F
ig
ur
e
4. G
r
a
phi
c
i
nt
e
r
f
a
c
e
of
t
he
a
s
s
is
ta
nc
e
pr
ot
ot
ype
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
. 14,
N
o. 5, Oc
to
be
r
2025
:
4202
-
4210
4206
O
nc
e
pow
e
r
e
d on, the
c
a
m
e
r
a
c
ont
in
uous
ly
m
oni
to
r
s
a
nd, upon
de
te
c
ti
ng t
ha
t
a
us
e
r
hol
ds
t
he
ir
ha
nd
up
f
or
a
t
le
a
s
t
2
-
s
e
c
onds
,
de
te
c
t
s
th
e
g
e
s
tu
r
e
a
nd
th
e
S
T
T
is
a
c
ti
va
te
d.
T
hi
s
pr
oc
e
s
s
e
n
s
ur
e
s
a
s
m
oot
h
tr
a
ns
it
io
n
be
twe
e
n
ge
s
tu
r
e
s
a
nd
s
ys
te
m
a
c
ti
va
ti
on.
O
nc
e
a
c
ti
v
a
te
d,
th
e
W
hi
s
pe
r
m
ode
l
tr
a
ns
c
r
ib
e
s
th
e
us
e
r
'
s
a
udi
o
in
put
in
r
e
a
l
ti
m
e
,
tr
a
ns
f
or
m
in
g
i
t
in
to
te
xt
w
it
h
hi
gh
a
c
c
ur
a
c
y.
T
hi
s
tr
a
ns
c
r
ip
ti
on
is
s
e
nt
di
r
e
c
tl
y
to
th
e
L
L
M
m
ode
l
f
o
r
pr
oc
e
s
s
in
g.
T
he
P
hi
-
3
M
in
i
m
ode
l
r
e
c
e
iv
e
s
th
e
tr
a
ns
c
r
ip
ti
on
a
nd
ge
ne
r
a
te
s
a
c
ohe
r
e
nt
a
nd
a
c
c
ur
a
te
r
e
s
pon
s
e
,
w
hi
c
h
is
di
s
pl
a
y
e
d
on
th
e
gr
a
phi
c
a
l
in
te
r
f
a
c
e
in
th
e
c
or
r
e
s
ponding
te
xt
box.
T
hi
s
c
om
pl
e
te
s
t
he
i
nt
e
r
a
c
ti
on c
yc
l
e
, a
ll
ow
in
g t
he
us
e
r
t
o r
e
c
e
iv
e
t
he
r
e
que
s
te
d i
nf
or
m
a
ti
on.
3.5.
V
al
id
at
io
n
O
nc
e
t
he
t
w
o m
ode
ls
a
r
e
e
m
be
dde
d i
n t
he
N
V
I
D
I
A
c
a
r
d, w
e
pr
oc
e
e
d t
o va
li
da
te
t
he
W
hi
s
pe
r
s
pe
e
c
h
r
e
c
ogni
ti
on
a
nd
s
pe
e
c
h
tr
a
ns
c
r
ip
ti
on
m
ode
l
a
nd
th
e
P
hi
-
3
M
in
i
na
tu
r
a
l
la
ngua
ge
m
ode
l.
F
or
th
e
W
hi
s
pe
r
m
ode
l,
10
va
li
da
ti
on
c
ont
e
xt
s
di
ve
r
s
if
ie
d
in
le
ngt
h,
li
ngui
s
ti
c
c
om
pl
e
xi
ty
a
nd
s
ty
le
a
r
e
e
s
ta
bl
is
he
d
in
c
lu
di
ng:
s
hor
t
a
nd
s
im
pl
e
te
xt
,
w
it
h
num
be
r
s
a
nd
da
ta
,
w
it
h
m
ul
ti
pl
e
a
c
c
e
nt
s
a
nd
punc
tu
a
ti
on
m
a
r
ks
,
w
it
h
in
f
or
m
a
l
la
ngua
ge
a
nd
c
ont
r
a
c
ti
ons
,
w
it
h
f
or
m
a
l
la
ngua
ge
a
nd
lo
ng
s
e
nt
e
nc
e
s
,
w
it
h
f
or
e
ig
n
w
or
ds
a
nd
m
ix
e
d
s
e
nt
e
nc
e
s
,
na
r
r
a
ti
ve
t
e
xt
w
it
h
de
s
c
r
ip
ti
ons
,
te
xt
w
it
h
di
a
lo
gue
b
e
twe
e
n
c
h
a
r
a
c
te
r
s
,
a
nd
te
xt
w
it
h
in
te
r
r
oga
ti
ve
a
nd
e
xc
la
m
a
to
r
y
s
e
nt
e
nc
e
s
.
F
or
e
a
c
h
of
th
e
m
,
bot
h
th
e
r
e
s
pons
e
ti
m
e
a
nd
th
e
le
ve
l
of
a
c
c
ur
a
c
y
in
th
e
tr
a
ns
c
r
ip
ti
on
(
in
S
pa
ni
s
h)
a
r
e
c
a
lc
ul
a
te
d,
w
hi
c
h
is
e
xpr
e
s
s
e
d
a
s
a
pe
r
c
e
nt
a
ge
of
th
e
r
a
ti
o
of
w
e
ll
-
id
e
nt
if
ie
d
c
ha
r
a
c
te
r
s
ove
r
th
e
to
ta
l
num
be
r
o
f
c
ha
r
a
c
te
r
s
(
w
it
hout
s
pa
c
e
)
of
e
a
c
h
te
s
t
s
e
nt
e
nc
e
;
T
a
bl
e
1
s
how
s
th
e
e
va
lu
a
ti
on of
t
he
s
pe
e
c
h t
r
a
n
s
c
r
ip
ti
on of
t
he
10 t
e
xt
s
t
ha
t
c
ove
r
e
d di
f
f
e
r
e
nt
s
c
e
na
r
io
s
, i
n
S
pa
ni
s
h
.
T
a
bl
e
1. L
oc
a
l
r
e
s
ul
ts
(
W
hi
s
pe
r
b
a
s
e
)
T
e
s
t
T
i
m
e
(
s
e
c
onds
)
A
c
c
ur
a
c
y
(%)
T
r
a
ns
c
r
i
be
d
t
e
xt
t
r
a
ns
c
r
i
be
d t
e
xt
w
i
t
h e
r
r
or
s
not
e
d
1
8.70
94.37
H
ol
a
, m
e
l
l
a
m
o j
ua
n y e
s
t
oy pr
oba
ndo
e
s
t
a
i
nt
e
l
i
ge
n
c
i
a
a
r
t
i
f
i
c
i
a
l
, _c
om
o e
s
t
a
s
h
oy_
2
14.43
99.02
E
l
pr
óxi
m
o
vue
l
o
a
N
ue
vo
Y
or
_
s
a
l
e
a
l
a
s
1545
de
s
de
l
a
pue
r
t
a
núm
e
r
o
27.
R
e
c
ue
r
da
que
e
l
núm
e
r
o de
c
onf
i
r
m
a
c
i
óne
s
A
B
123456.
3
11.81
85.71
L
a
c
onf
e
r
e
nc
i
a
s
obr
e
l
a
e
xpe
r
i
_e
nc
i
a
a
r
t
i
f
i
c
i
a
l
s
e
l
e
i
m
pa
r
c
í
_a
por
l
a
_O
C
T
O
R
E
S
M
I
G
y l
a
_O
C
T
O
R
E
S
G
a
r
c
í
a
e
n l
a
O
ul
a
101 __, l
a
U
ni
ve
r
s
i
da
d
T
e
c
nol
ógi
c
a
.
4
10.89
90.43
H
ol
a
,
_C
om
o
e
s
t
a
s
_
A
ye
r
vi
a
m
e
m
i
gope
r
o
e
n
l
a
t
i
e
nda
y
v
a
ya
s
or
pr
e
s
a
,
t
a
m
bi
é
n
e
s
t
a
ba
a
na
, __
c
_a
s
i
a
t
i
e
m
po __ ve
i
a
5
9.9
93.02
N
o
t
e
pr
oc
u_pe
s
que
y
a
t
e
ngo
t
odo
l
i
s
t
o
v
a
m
os
a
ve
r
s
i
l
a
__
e
s
t
a
e
l
a
i
r
e
r
e
c
on
oc
e
bi
e
n
t
odo l
o que
di
go va
l
e
6
11.46
94.90
C
on
e
l
pr
e
s
e
nt
e
doc
um
e
nt
o
m
e
p
e
r
m
i
t
e
i
nf
or
m
a
r
e
l
e
s
c
r
a
pe
r
pr
oye
c
t
o
ha
s
i
do
c
om
pl
e
t
a
dos
s
a
t
i
s
f
a
c
t
or
i
a
m
e
nt
e
,
c
um
pl
i
e
ndo c
on
t
odos
l
os
p
a
r
á
m
e
t
r
os
e
s
t
a
bl
e
c
i
dos
c
on
e
l
c
ont
r
a
t
o or
i
gi
na
l
7
9.35
100.00
E
l
nue
vo
s
of
t
w
a
r
e
ut
i
l
i
z
a
un
a
l
gor
i
t
m
o
ba
s
a
do
e
n
m
a
c
hi
ne
l
e
a
r
ni
ng
que
m
e
j
or
a
e
l
r
e
ndi
m
i
e
nt
o de
l
s
i
s
t
e
m
a
de
r
e
c
onoc
i
m
i
e
nt
o de
voz
8
11.23
99.12
E
l
s
ol
s
e
oc
ul
t
a
ba
ha
s
t
a
l
a
m
ont
a
ña
m
i
e
nt
r
a
s
e
l
vi
e
nt
o
a
c
a
r
i
c
i
a
ba
s
ua
ve
m
e
nt
e
l
a
s
s
oj
a
s
de
l
os
a
r
bol
e
s
. E
r
a
U
na
t
a
r
de
t
r
a
nqui
l
a
e
n pe
que
9
8.67
100.00
¿
E
s
t
á
s
s
e
gur
o de
que
e
s
una
bu
e
na
i
de
a
?
---
pr
e
gunt
ó A
na
.
---
S
í
, c
onf
í
a
e
n m
í
. H
e
he
c
ho e
s
t
o m
uc
ha
s
ve
c
e
s
a
nt
e
s
10
8.61
90.91
Q
ui
e
r
_o_ve
r
i
m
a
gi
na
do que
t
odo t
e
r
m
i
na
r
a
s
i
e
s
i
nc
r
e
í
bl
e
l
o qu
e
ha
s
uc
e
do e
n
t
a
n poc
o
t
i
e
m
po
F
r
om
th
e
t
e
s
t
s
pe
r
f
or
m
e
d
,
th
e
a
ve
r
a
g
e
r
e
c
ogn
it
i
on
ti
m
e
f
or
e
a
c
h
te
xt
i
s
10
.5
05
s
e
c
o
nd
s
,
w
hi
c
h
i
s
e
qu
iv
a
l
e
nt
t
o
0.
108
s
e
c
ond
s
p
e
r
c
h
a
r
a
c
t
e
r
,
a
nd
t
he
a
v
e
r
a
g
e
pe
r
c
e
nt
a
g
e
of
a
c
c
ur
a
c
y
in
th
e
tr
a
n
s
c
r
ip
ti
o
n
of
th
e
s
p
e
e
c
h i
s
94
.75
%
. I
n
a
d
d,
k
e
y
pa
r
a
m
e
t
e
r
s
s
uc
h
a
s
a
ud
io
qu
a
li
ty
,
pl
a
yb
a
c
k f
lu
e
n
c
y,
a
nd
f
r
a
gm
e
n
t
s
y
nc
hr
o
ni
z
a
t
io
n
w
e
r
e
e
v
a
lu
a
t
e
d
. T
h
e
r
e
s
ul
t
s
d
e
m
o
n
s
tr
a
t
e
d c
o
ns
i
s
te
nt
pe
r
f
or
m
a
n
c
e
, c
h
a
r
a
c
te
r
iz
e
d by mi
n
im
a
l
r
e
s
po
n
s
e
t
im
e
s
a
nd
a
ud
io
qu
a
l
it
y
s
ui
t
a
b
le
f
or
n
a
t
ur
a
l
i
nt
e
r
a
c
t
io
n
s
,
a
lt
hou
gh
s
li
g
ht
p
a
u
s
e
s
w
e
r
e
i
de
nt
if
ie
d i
n
s
o
m
e
f
r
a
gm
e
n
ts
.
T
he
r
e
f
or
e
,
it
is
pos
s
ib
le
to
a
f
f
ir
m
th
a
t
th
e
m
ode
l
pr
ove
d
to
be
e
f
f
ic
ie
nt
,
w
it
h
hi
gh
a
c
c
ur
a
c
y
a
nd
no
a
s
s
oc
ia
t
e
d
c
os
t
due
to
it
s
lo
c
a
l
e
xe
c
ut
io
n.
I
t
is
s
ui
ta
bl
e
f
or
r
e
a
l
-
ti
m
e
tr
a
ns
c
r
ip
ti
on
a
nd
va
r
ie
d
s
c
e
na
r
io
s
,
m
a
tc
hi
ng
or
s
ur
pa
s
s
in
g
s
im
il
a
r
m
ode
ls
in
lo
c
a
l
ha
r
dw
a
r
e
e
nvi
r
onm
e
nt
s
.
F
or
th
e
P
hi
-
3
M
in
i
m
ode
l
,
10
s
c
e
na
r
io
s
w
e
r
e
de
f
in
e
d
w
it
h
que
s
ti
ons
or
in
s
tr
uc
ti
ons
c
ove
r
in
g
ge
ne
r
a
l
a
nd
s
pe
c
if
ic
to
pi
c
s
,
f
or
e
a
c
h
th
e
a
c
c
ur
a
c
y a
nd
c
ohe
r
e
n
c
e
of
th
e
a
ns
w
e
r
gi
ve
n
w
a
s
e
va
lu
a
te
d
on a
s
c
a
le
of
1
to
10,
w
he
r
e
10
c
or
r
e
s
pond
s
to
th
e
be
s
t
s
c
or
e
.
T
a
bl
e
2
s
how
s
t
he
e
v
a
lu
a
ti
on of
t
he
a
c
c
ur
a
c
y
a
nd
c
o
ns
is
te
nc
y
of
t
he
10 s
c
e
na
r
io
s
.
W
it
h
a
n
a
ve
r
a
ge
a
c
c
ur
a
c
y
of
9.5/
10
a
nd
a
n
a
v
e
r
a
ge
c
on
s
is
te
nc
y
of
8.5/
10,
P
hi
-
3
M
in
i
pr
ovi
de
d
c
ons
is
te
nt
a
nd
a
c
c
ur
a
te
a
n
s
w
e
r
s
,
s
t
a
ndi
ng
out
f
or
it
s
s
pe
e
d
a
nd
lo
w
r
e
s
our
c
e
c
ons
um
pt
io
n.
S
in
c
e
it
r
uns
lo
c
a
ll
y,
it
doe
s
not
ge
ne
r
a
te
us
a
ge
c
os
t
s
,
w
hi
c
h
m
a
ke
s
it
a
n
e
c
o
nom
ic
a
l
a
nd
e
f
f
ic
ie
nt
s
ol
ut
io
n.
T
he
in
te
gr
a
te
d
w
or
kf
lo
w
de
s
ig
ne
d
e
ns
ur
e
s
e
f
f
ic
ie
nt
(
F
ig
ur
e
5)
,
in
tu
i
ti
ve
a
nd
a
da
pt
iv
e
in
te
r
a
c
ti
ve
e
xpe
r
ie
nc
e
,
l
e
ve
r
a
gi
ng
a
dva
nc
e
d
ge
s
tu
r
e
de
te
c
ti
on
a
nd
la
ngua
ge
pr
oc
e
s
s
in
g
c
a
pa
bi
li
ti
e
s
in
a
n
opt
im
iz
e
d
lo
c
a
l
e
nvi
r
onm
e
nt
.
F
in
a
ll
y,
a
s
s
how
n
in
F
ig
ur
e
4,
th
e
pe
r
f
or
m
a
nc
e
of
th
e
opt
im
iz
e
d
gr
a
phi
c
in
te
r
f
a
c
e
is
va
li
da
t
e
d
by
r
a
is
in
g
th
e
ha
nd
a
nd
ve
r
ba
ll
y
f
or
m
ul
a
ti
ng
th
e
que
s
ti
on
“
W
ha
t
i
s
th
e
c
lo
s
e
s
t
h
a
bi
ta
b
le
pl
a
ne
t
to
e
a
r
th
?
”
a
nd
qui
c
kl
y
obt
a
in
in
g
a
n
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
L
e
ar
ni
ng as
s
i
s
ta
nc
e
m
odul
e
ba
s
e
d on a s
m
al
l
la
nguage
m
od
e
l
(
M
ar
c
o A
nt
oni
o J
in
e
te
)
4207
a
c
c
ur
a
te
a
nd
c
ohe
r
e
nt
a
ns
w
e
r
,
th
e
te
s
t
pe
r
f
or
m
a
nc
e
r
e
s
ul
t
s
h
ow
s
th
a
t
th
e
in
te
gr
a
te
d
w
or
kf
lo
w
e
ns
ur
e
s
a
n
e
f
f
ic
ie
nt
,
in
tu
it
iv
e
a
nd
a
da
pt
iv
e
in
te
r
a
c
ti
ve
e
xpe
r
ie
nc
e
,
ta
ki
ng
a
dva
nt
a
ge
of
a
dva
nc
e
d
ge
s
tu
r
e
de
te
c
ti
on
a
nd
la
ngua
ge
pr
oc
e
s
s
in
g c
a
pa
bi
li
ti
e
s
i
n
a
n opti
m
iz
e
d l
oc
a
l
e
nvi
r
on
m
e
nt
.
T
a
bl
e
2. L
oc
a
l
r
e
s
ul
ts
(
P
hi
-
3 M
in
i
m
ode
l)
T
e
s
t
Q
ue
s
t
i
on/
i
ns
t
r
uc
t
i
on
A
ns
w
e
r
A
c
c
ur
a
c
y
C
ons
i
s
t
e
nc
y
1
W
ha
t
i
s
t
he
c
a
pi
t
a
l
of
F
r
a
nc
e
?
P
a
r
i
s
.
10
10
2
T
he
gl
a
s
s
i
s
f
ul
l
of
w
a
t
e
r
,
w
ha
t
ha
ppe
ns
i
f
i
t
br
e
a
ks
?
I
f
a
gl
a
s
s
f
ul
l
of
w
a
t
e
r
br
e
a
ks
,
i
t
i
s
l
i
ke
l
y
t
ha
t
m
os
t
of
t
he
c
ont
e
nt
s
w
i
l
l
be
s
pi
l
l
e
d
a
l
l
ove
r
t
he
pl
a
c
e
.
T
he
e
x
a
c
t
a
m
ount
of
w
a
t
e
r
l
os
t
w
i
l
l
de
pe
nd on s
e
ve
r
a
l
f
a
c
t
or
s
...
10
10
3
H
ow
m
uc
h i
s
4+5+
10+8+
9+3+
4?
4+5+
10+8+
9+3+
4=42
.
T
he
r
e
f
or
e
,
t
he
a
ns
w
e
r
i
s
42
10
10
4
W
ha
t
da
y
i
s
t
oda
y
a
nd
w
ha
t
w
i
l
l
t
he
w
e
a
t
he
r
be
t
om
or
r
ow
?
T
oda
y
i
s
W
e
dne
s
da
y,
M
a
r
c
h
2023.
T
om
or
r
ow
t
he
w
e
a
t
he
r
w
i
l
l
be
s
unny
w
i
t
h
a
n
a
ve
r
a
ge
t
e
m
pe
r
a
t
ur
e
of
15 C
e
l
s
i
us
.
10
0
5
H
ow
do you s
a
y he
l
l
o i
n E
ngl
i
s
h?
H
e
l
l
o
10
10
6
H
ow
doe
s
c
l
i
m
a
t
e
c
ha
ng
e
a
f
f
e
c
t
gl
a
c
i
e
r
s
?
C
l
i
m
a
t
e
c
ha
nge
ha
s
a
di
r
e
c
t
a
nd
i
m
pa
c
t
i
ng
i
nf
l
ue
nc
e
on
gl
a
c
i
e
r
dyna
m
i
c
s
by
a
c
c
e
l
e
r
a
t
i
ng
or
s
l
ow
i
ng
t
he
i
r
m
ove
m
e
nt
or
c
ont
r
i
but
i
ng
t
o
f
ur
t
he
r
m
e
l
t
i
ng.
T
he
y
a
f
f
e
c
t
m
a
i
nl
y
t
h
r
ough
t
w
o
pa
t
hw
a
ys
.
G
l
oba
l
i
nc
r
e
a
s
e
a
nd m
i
ni
m
um
w
i
nt
e
r
i
c
e
...
10
10
7
E
xpl
a
i
n
t
o
m
e
i
n
t
hr
e
e
s
t
e
ps
how
t
o pr
e
pa
r
e
a
c
of
f
e
e
?
t
o
m
a
ke
a
s
i
m
pl
e
but
t
a
s
t
y
c
up
of
c
of
f
e
e
w
i
t
h
boi
l
i
ng
w
a
t
e
r
a
nd
gr
ound
be
a
ns
(
i
.e
.,
not
us
i
ng
s
pe
c
i
a
l
i
z
e
d
a
ppl
i
a
nc
e
s
s
u
c
h
a
s
e
s
pr
e
s
s
o
m
a
ke
r
s
or
t
ur
bo
m
a
c
hi
ne
s
)
.
F
ol
l
ow
t
he
s
e
t
hr
e
e
s
t
e
ps
...
10
10
8
W
ha
t
do
you
t
hi
nk
a
bout
t
he
l
a
t
e
s
t
a
r
t
i
f
i
c
i
a
l
i
nt
e
l
l
i
ge
nc
e
t
e
c
hnol
ogy?
C
ur
r
e
nt
l
y
t
he
l
a
t
e
s
t
a
dva
nc
e
d
t
e
c
hnol
ogy
i
n
A
I
,
i
s
m
a
i
nl
y
f
oc
us
e
d
on
i
m
p
r
ovi
ng
i
t
s
na
t
u
r
a
l
l
a
ngua
ge
unde
r
s
t
a
ndi
ng
(
N
L
U
)
c
a
pa
bi
l
i
t
i
e
s
,
i
m
pr
ovi
ng
t
r
a
ns
f
e
r
l
e
a
r
ni
ng
a
nd
a
ut
onom
ous
a
ppl
i
c
a
t
i
ons
w
i
t
h
a
hi
ghe
r
de
gr
e
e
of
a
ut
onom
y....
10
10
9
R
e
pe
a
t
w
ha
t
I
j
us
t
s
a
i
d
I
unde
r
s
t
a
nd.
A
s
a
h
e
l
pf
ul
a
s
s
i
s
t
a
nt
,
m
y
goa
l
i
s
t
o
pr
ovi
de
e
f
f
i
c
i
e
nt
a
nd
a
c
c
ur
a
t
e
s
ol
ut
i
ons
w
i
t
h
s
pe
e
d
a
nd
opt
i
m
a
l
i
t
y t
o m
a
xi
m
i
z
e
us
e
r
s
a
t
i
s
f
a
c
t
i
on
5
5
10
W
ha
t
do
you
t
hi
nk
a
bout
s
ha
r
i
ng
ba
nki
ng da
t
a
on t
he
i
nt
e
r
ne
t
?
S
ha
r
i
ng
ba
nki
ng
da
t
a
on
t
he
i
nt
e
r
ne
t
c
a
n
be
a
r
i
s
ky
a
c
t
i
vi
t
y w
i
t
hout
due
c
onc
e
r
n f
or
s
e
c
ur
i
t
y...
10
10
F
ig
ur
e
5. G
r
a
phi
c
i
nt
e
r
f
a
c
e
of
t
he
a
s
s
is
ta
nc
e
pr
ot
ot
ype
I
n
ge
ne
r
a
l,
th
e
g
r
a
phi
c
a
l
in
te
r
f
a
c
e
de
ve
lo
pe
d
f
or
th
is
pr
o
je
c
t
e
f
f
ic
ie
nt
ly
in
te
gr
a
te
s
th
e
f
unc
ti
ona
li
t
ie
s
of
th
e
W
hi
s
p
e
r
(
S
T
T
)
a
nd
P
hi
-
3
M
in
i
(
L
L
M
)
m
ode
ls
in
to
a
n
i
nt
e
r
a
c
ti
ve
e
nvi
r
onm
e
nt
r
unni
ng
lo
c
a
ll
y
on
th
e
N
V
I
D
I
A
J
e
ts
on
X
a
vi
e
r
A
G
X
.
T
hi
s
c
om
pon
e
nt
a
c
ts
a
s
th
e
c
or
e
of
in
te
r
a
c
ti
on
be
twe
e
n
th
e
us
e
r
a
nd
th
e
s
ys
te
m
,
a
ll
ow
in
g
voi
c
e
tr
a
ns
c
r
ip
ti
on,
r
e
s
pons
e
ge
ne
r
a
ti
on
a
nd
a
c
ti
va
ti
on
th
r
ough
g
e
s
tu
r
e
de
t
e
c
ti
on.
A
ddi
ti
ona
ll
y,
it
is
im
por
ta
nt
to
not
e
th
a
t
th
e
pe
r
f
or
m
a
nc
e
of
th
e
boa
r
d
is
s
ig
ni
f
ic
a
nt
on
a
ll
c
or
e
s
a
t
th
e
s
ta
r
t
of
L
L
M
a
s
s
how
n i
n F
ig
ur
e
6.
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
. 14,
N
o. 5, Oc
to
be
r
2025
:
4202
-
4210
4208
F
ig
ur
e
6. N
V
I
D
I
A
J
e
ts
on
X
a
vi
e
r
c
a
r
d pe
r
f
or
m
a
nc
e
4.
C
O
N
C
L
U
S
I
O
N
T
he
de
ve
lo
pm
e
nt
a
nd
e
va
lu
a
ti
on
of
th
e
W
hi
s
pe
r
(
m
ode
l
ba
s
e
)
a
nd
P
hi
-
3
M
in
i
(
3
.
8B
)
m
ode
ls
ha
ve
de
m
ons
tr
a
te
d
th
a
t
it
is
po
s
s
ib
le
to
im
pl
e
m
e
nt
e
f
f
ic
ie
nt
,
a
c
c
ur
a
te
a
nd
c
os
t
-
e
f
f
e
c
ti
ve
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
s
ol
ut
io
ns
on
lo
c
a
l
ha
r
dw
a
r
e
de
vi
c
e
s
, s
uc
h
a
s
th
e
N
V
I
D
I
A
J
e
t
s
o
n
X
a
vi
e
r
A
G
X
.
T
he
r
e
s
ul
ts
obt
a
in
e
d
hi
ghl
ig
ht
th
e
a
bi
li
ty
of
bot
h m
ode
ls
t
o m
e
e
t
th
e
r
e
qui
r
e
m
e
nt
s
of
s
pe
e
c
h t
r
a
ns
c
r
ip
ti
on a
nd l
a
ngua
ge
ge
ne
r
a
ti
on w
it
h hi
gh
le
ve
ls
of
a
c
c
ur
a
c
y,
c
ons
i
s
te
nc
y
a
nd
s
pe
e
d,
w
it
hout
in
c
ur
r
in
g
c
os
ts
a
s
s
o
c
ia
te
d
w
it
h
th
e
ir
e
xe
c
ut
io
n.
F
ur
th
e
r
m
or
e
,
th
e
in
te
gr
a
ti
on
of
th
e
s
e
m
ode
ls
in
to
a
gr
a
phi
c
a
l
in
te
r
f
a
c
e
w
it
h
ge
s
tu
r
e
de
te
c
ti
on
e
nha
nc
e
s
hum
a
n
-
m
a
c
hi
ne
in
te
r
a
c
ti
on
by
pr
ovi
di
ng
a
n
in
tu
it
iv
e
a
nd
dyna
m
ic
e
xpe
r
ie
nc
e
.
T
hi
s
s
ys
te
m
not
onl
y
opt
im
iz
e
s
th
e
us
e
of
r
e
s
our
c
e
s
but
a
ls
o
ope
ns
ne
w
pos
s
ib
il
it
ie
s
f
or
it
s
a
ppl
ic
a
ti
on
in
a
c
c
e
s
s
ib
le
,
in
te
r
a
c
ti
ve
a
nd
a
ut
onomous
e
nvi
r
onm
e
nt
s
,
m
a
r
ki
ng
a
s
ig
ni
f
ic
a
nt
a
dv
a
nc
e
t
ow
a
r
ds
m
or
e
r
obus
t
hum
a
n
-
m
a
c
hi
ne
in
te
r
f
a
c
e
s
ol
ut
io
ns
a
da
pt
a
bl
e
to
di
v
e
r
s
e
ne
e
ds
.
T
he
in
te
gr
a
ti
on
of
th
e
T
T
S
m
odul
e
in
to
th
e
ove
r
a
ll
s
ys
te
m
r
e
pr
e
s
e
nt
s
a
s
ig
ni
f
ic
a
nt
a
dva
nc
e
in
th
e
d
e
ve
lo
pm
e
nt
of
in
te
r
a
c
ti
ve
in
te
r
f
a
c
e
s
ba
s
e
d
on
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
. T
he
c
hunking
s
tr
a
te
gy
not
onl
y
e
nha
nc
e
s
s
ys
te
m
e
f
f
ic
ie
nc
y
but
a
ls
o
e
ns
ur
e
s
opt
im
a
l
us
e
r
e
xpe
r
ie
nc
e
by
r
e
duc
in
g
w
a
it
in
g
ti
m
e
s
a
nd
m
a
in
ta
in
in
g
r
e
s
pons
e
f
lu
id
it
y.
T
hi
s
a
ppr
oa
c
h
a
li
gns
w
it
h
th
e
pr
oj
e
c
t’
s
obj
e
c
ti
ve
s
by
gua
r
a
nt
e
e
in
g
in
tu
it
iv
e
, dyna
m
ic
, a
nd a
c
c
e
s
s
ib
le
i
nt
e
r
a
c
ti
on f
or
us
e
r
s
i
n l
oc
a
l
e
nvi
r
onm
e
nt
s
w
it
h opti
m
iz
e
d ha
r
dw
a
r
e
.
A
C
K
N
O
WL
E
D
G
E
M
E
N
T
S
T
he
a
ut
hor
s
e
xpr
e
s
s
th
e
ir
gr
a
ti
tu
de
to
U
ni
ve
r
s
id
a
d
M
il
it
a
r
N
ue
va
G
r
a
na
da
w
he
r
e
th
e
y
a
r
e
f
ul
l
ti
m
e
pr
of
e
s
s
or
s
,
a
nd
to
C
é
s
a
r
A
ugus
to
C
a
ld
e
r
ón,
a
m
e
c
ha
tr
oni
c
s
e
ngi
ne
e
r
in
g
s
tu
de
nt
,
f
or
hi
s
c
ont
r
ib
ut
io
n
to
th
e
de
ve
lo
pm
e
nt
of
t
he
i
nt
e
r
f
a
c
e
a
nd t
he
i
ns
ta
ll
a
ti
on of
t
ool
s
f
or
t
hi
s
r
e
s
e
a
r
c
h.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
P
r
oduc
t
de
r
iv
e
d
f
r
om
th
e
r
e
s
e
a
r
c
h
pr
oj
e
c
t
ti
tl
e
d
“
D
is
e
ño
de
un
m
ode
lo
de
in
te
r
a
c
c
ió
n
hum
a
no
-
r
obot
m
e
di
a
nt
e
a
lg
or
it
m
os
de
a
pr
e
ndi
z
a
je
pr
of
undo”
I
N
V
-
I
N
G
-
397
1.
P
r
oj
e
c
t
f
unde
d
by
th
e
O
f
f
ic
e
of
th
e
V
ic
e
P
r
e
s
id
e
nt
f
or
R
e
s
e
a
r
c
h a
t
th
e
U
ni
ve
r
s
id
a
d M
il
it
a
r
N
ue
va
G
r
a
na
da
f
or
t
he
f
is
c
a
l
ye
a
r
2024.
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
d
iT
)
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.
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
L
e
ar
ni
ng as
s
i
s
ta
nc
e
m
odul
e
ba
s
e
d on a s
m
al
l
la
nguage
m
od
e
l
(
M
ar
c
o A
nt
oni
o J
in
e
te
)
4209
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
M
a
r
c
o A
nt
oni
o J
in
e
te
✓
✓
✓
✓
✓
R
obi
ns
on
J
im
é
n
e
z
-
M
or
e
no
✓
✓
✓
✓
✓
✓
✓
✓
A
nny E
s
pi
ti
a
-
C
ubi
ll
os
✓
✓
✓
✓
✓
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
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
D
a
t
a
a
va
i
la
bi
l
it
y
i
s
not
a
pp
li
c
a
b
le
t
o t
hi
s
pa
pe
r
a
s
no
n
e
w
d
a
t
a
w
e
r
e
c
r
e
a
t
e
d
or
a
na
ly
z
e
d
in
th
i
s
s
tu
d
y.
R
E
F
E
R
E
N
C
E
S
[
1]
Y
.
L
i
u
e
t
al
.
,
“
U
nde
r
s
t
a
ndi
ng
L
L
M
s
:
a
c
om
pr
e
he
ns
i
v
e
ove
r
vi
e
w
f
r
om
t
r
a
i
ni
n
g
t
o
i
nf
e
r
e
nc
e
,”
N
e
ur
oc
om
put
i
ng
,
vol
.
620,
2025,
doi
:
10.1016/
j
.ne
uc
om
.2024.129190.
[
2]
Y
.
S
un
e
t
al
.
,
“
D
e
v
e
l
opm
e
nt
of
a
n
i
nt
e
l
l
i
ge
nt
de
s
i
gn
a
nd
s
i
m
ul
a
t
i
on
a
i
d
s
ys
t
e
m
f
or
he
a
t
t
r
e
a
t
m
e
nt
pr
oc
e
s
s
e
s
b
a
s
e
d
on
L
L
M
,
”
M
at
e
r
i
al
s
and D
e
s
i
gn
, vol
. 248, 2024, doi
:
10.1016/
j
.m
a
t
de
s
.2024.113506.
[
3]
R
.
Z
a
he
di
f
a
r
,
M
.
S
.
B
a
ghs
ha
h,
a
nd
A
.
T
a
he
r
i
,
“
L
L
M
-
c
ont
r
ol
l
e
r
:
dyna
m
i
c
r
obo
t
c
ont
r
ol
a
da
pt
a
t
i
on
us
i
ng
l
a
r
ge
l
a
ngua
ge
m
ode
l
s
,”
R
obot
i
c
s
and A
ut
onom
ous
Sy
s
t
e
m
s
, vol
. 186, A
pr
. 2025, doi
:
10.1016/
j
.r
obot
.2024.104913.
[
4]
A
.
M
a
hm
ood,
J
.
W
a
ng,
B
.
Y
a
o,
D
.
W
a
ng,
a
nd
C
.
M
.
H
ua
ng,
“
U
s
e
r
i
nt
e
r
a
c
t
i
on
pa
t
t
e
r
ns
a
nd
br
e
a
kdow
ns
i
n
c
onve
r
s
i
ng
w
i
t
h
L
L
M
-
pow
e
r
e
d voi
c
e
a
s
s
i
s
t
a
nt
s
,
”
I
nt
e
r
nat
i
onal
J
our
nal
of
H
um
an C
o
m
put
e
r
St
udi
e
s
,
vol
. 195, 2025, doi
:
10.1016/
j
.i
j
hc
s
.2024.103406.
[
5]
C
.
I
.
G
a
r
c
i
a
,
M
.
A
.
D
i
B
a
t
t
i
s
t
a
,
T
.
A
.
L
e
t
e
l
i
e
r
,
H
.
D
.
H
a
l
l
or
a
n,
a
nd
J
.
A
.
C
a
m
e
l
i
o,
“
F
r
a
m
e
w
or
k
f
or
L
L
M
a
ppl
i
c
a
t
i
ons
i
n
m
a
nuf
a
c
t
ur
i
ng,”
M
anuf
ac
t
ur
i
ng L
e
t
t
e
r
s
, vol
. 41, pp. 253
–
263, 2024, doi
:
10.1016/
j
.m
f
gl
e
t
.2024.09.030.
[
6]
X
.
L
i
u,
J
.
A
.
E
r
k
oy
unc
u,
J
.
Y
.
H
.
F
u
h,
W
.
F
.
L
u
,
a
n
d
B
.
L
i
,
“
K
n
ow
l
e
dg
e
e
xt
r
a
c
t
i
on
f
or
a
d
di
t
i
ve
m
a
n
uf
a
c
t
ur
i
n
g
pr
oc
e
s
s
vi
a
na
m
e
d
e
nt
i
t
y
r
e
c
o
gn
i
t
i
o
n w
i
t
h L
L
M
s
,”
R
ob
ot
i
c
s
an
d C
o
m
pu
t
e
r
-
I
n
t
e
gr
a
t
e
d
M
a
nuf
ac
t
ur
i
n
g
,
vo
l
.
93
, 2
02
5,
do
i
:
10
.1
016
/
j
.
r
c
i
m
.
202
4.
102
90
0.
[
7]
M
.
L
.
T
s
a
i
,
C
.
W
.
O
ng,
a
nd
C
.
L
.
C
he
n,
“
E
xpl
or
i
ng
t
he
us
e
of
l
a
r
ge
l
a
ngua
ge
m
ode
l
s
(
L
L
M
s
)
i
n
c
he
m
i
c
a
l
e
ngi
ne
e
r
i
ng
e
duc
a
t
i
on
:
bui
l
di
ng
c
or
e
c
our
s
e
pr
obl
e
m
m
ode
l
s
w
i
t
h
C
ha
t
-
G
P
T
,”
E
duc
at
i
on
f
or
C
he
m
i
c
al
E
ngi
ne
e
r
s
,
vol
.
44,
pp.
71
–
95,
2023
,
doi
:
10.1016/
j
.e
c
e
.2023.05.001.
[
8]
S
.
P
a
ga
no
e
t
al
.
,
“
E
va
l
ua
t
i
ng
C
ha
t
G
P
T
,
G
e
m
i
ni
a
nd
ot
he
r
l
a
r
ge
l
a
ngua
g
e
m
ode
l
s
(
L
L
M
s
)
i
n
or
t
hopa
e
di
c
di
a
gnos
t
i
c
s
:
a
pr
os
pe
c
t
i
ve
c
l
i
ni
c
a
l
s
t
udy,”
C
om
put
at
i
onal
and
St
r
uc
t
ur
al
B
i
ot
e
c
hnol
ogy
J
our
nal
,
vol
.
28,
pp.
9
–
15,
2025,
doi
:
10.1016/
j
.c
s
bj
.2024.12.013.
[
9]
F
.
X
u
,
T
.
Z
ho
u,
T
.
N
g
uye
n,
H
.
B
a
o,
C
.
L
i
n,
a
nd
J
.
D
u
,
“
I
nt
e
g
r
a
t
i
n
g
a
ug
m
e
nt
e
d
r
e
a
l
i
t
y
a
nd
L
L
M
f
or
e
nha
nc
e
d
c
og
ni
t
i
ve
s
up
po
r
t
i
n
c
r
i
t
i
c
a
l
a
u
di
o c
om
m
u
ni
c
a
t
i
ons
,”
I
n
t
e
r
n
at
i
o
na
l
J
o
ur
n
al
of
H
um
a
n C
om
p
ut
e
r
S
t
ud
i
e
s
, v
ol
.
1
94,
2
025
,
do
i
:
10
.10
16
/
j
.i
j
hc
s
.20
24
.10
34
02.
[
10]
Z
.
D
e
ng
e
t
al
.
,
“
O
phG
L
M
:
a
n
opht
ha
l
m
ol
ogy
l
a
r
ge
l
a
ngua
ge
-
a
nd
-
vi
s
i
on
a
s
s
i
s
t
a
nt
,”
A
r
t
i
f
i
c
i
al
I
n
t
e
l
l
i
ge
nc
e
i
n
M
e
di
c
i
ne
,
vol
.
157,
2024, doi
:
10.1016/
j
.a
r
t
m
e
d.2024.103001.
[
11]
J
. S
. E
r
i
c
ks
on, H
. S
a
nt
os
, V
. P
i
nhe
i
r
o, J
. P
. M
c
C
us
ke
r
, a
nd D
.
L
.
M
c
G
ui
nne
s
s
,
“
L
L
M
e
xpe
r
i
m
e
nt
a
t
i
on t
hr
ough know
l
e
dge
gr
a
phs
:
t
ow
a
r
ds
i
m
pr
ove
d
m
a
na
ge
m
e
nt
,
r
e
pe
a
t
a
bi
l
i
t
y,
a
nd
ve
r
i
f
i
c
a
t
i
on,”
J
our
n
al
of
W
e
b
Se
m
ant
i
c
s
,
vol
.
85,
2025,
doi
:
10.1016/
j
.w
e
bs
e
m
.2024.100853.
[
12]
M
.
G
a
r
r
y,
W
.
M
.
C
ha
n,
J
.
F
os
t
e
r
,
a
nd
L
.
A
.
H
e
nke
l
,
“
L
a
r
ge
l
a
ngua
g
e
m
ode
l
s
(
L
L
M
s
)
a
nd
t
he
i
ns
t
i
t
ut
i
ona
l
i
z
a
t
i
on
of
m
i
s
i
nf
or
m
a
t
i
on,”
T
r
e
nds
i
n C
ogni
t
i
v
e
Sc
i
e
nc
e
s
, vol
. 28, no. 12, pp. 1078
–
1088,
2024, doi
:
10.1016/
j
.t
i
c
s
.2024.08.007.
[
13]
S
.
S
ha
f
e
e
,
A
.
B
e
s
s
a
ni
,
a
nd
P
.
M
.
F
e
r
r
e
i
r
a
,
“
E
va
l
ua
t
i
on
of
L
L
M
-
ba
s
e
d
c
ha
t
bo
t
s
f
or
os
i
nt
-
ba
s
e
d
c
yb
e
r
t
hr
e
a
t
a
w
a
r
e
ne
s
s
,”
E
x
pe
r
t
Sy
s
t
e
m
s
w
i
t
h A
ppl
i
c
at
i
ons
, vol
. 261, 2025, doi
:
10.1016/
j
.e
s
w
a
.2024.125509.
[
14]
J
.
W
e
s
t
e
r
,
S
.
de
J
ong,
H
.
P
ohl
,
a
nd
N
.
va
n
B
e
r
ke
l
,
“
E
xpl
or
i
ng
pe
opl
e
’
s
pe
r
c
e
pt
i
ons
of
L
L
M
-
ge
ne
r
a
t
e
d
a
dvi
c
e
,”
C
om
put
e
r
s
i
n
H
um
an B
e
hav
i
or
:
A
r
t
i
f
i
c
i
al
H
um
ans
, vol
. 2, no. 2, 2024, doi
:
10.1016/
j
.c
hba
h.2024.100072.
[
15]
J
.
R
obe
r
t
s
,
M
.
B
a
ke
r
,
a
nd
J
.
A
ndr
e
w
,
“
A
r
t
i
f
i
c
i
a
l
i
nt
e
l
l
i
ge
nc
e
a
nd
qua
l
i
t
a
t
i
ve
r
e
s
e
a
r
c
h:
t
he
pr
om
i
s
e
a
nd
pe
r
i
l
s
of
l
a
r
ge
l
a
ngu
a
g
e
m
ode
l
(
L
L
M
)
‘
a
s
s
i
s
t
a
nc
e
,’
”
C
r
i
t
i
c
al
P
e
r
s
pe
c
t
i
v
e
s
on A
c
c
ount
i
ng
, vol
. 99, M
a
r
.
2024, doi
:
10.1016/
j
.c
pa
.2024.102722.
[
16]
B
.
A
l
s
a
f
a
r
i
,
E
.
A
t
w
e
l
l
,
A
.
W
a
l
ke
r
,
a
nd
M
.
C
a
l
l
a
gha
n,
“
T
ow
a
r
ds
e
f
f
e
c
t
i
ve
t
e
a
c
h
i
ng
a
s
s
i
s
t
a
nt
s
:
f
r
om
i
nt
e
nt
-
ba
s
e
d
c
ha
t
bot
s
t
o
L
L
M
-
pow
e
r
e
d t
e
a
c
hi
ng a
s
s
i
s
t
a
nt
s
,
”
N
at
ur
al
L
anguage
P
r
oc
e
s
s
i
ng J
our
nal
, vol
. 8, p.
100101, 2024, doi
:
10.1016/
j
.nl
p.2024.100101.
[
17]
S
.
A
l
s
a
qe
r
,
S
.
A
l
a
j
m
i
,
I
.
A
hm
a
d,
a
nd
M
.
A
l
f
a
i
l
a
ka
w
i
,
“
T
he
pot
e
nt
i
a
l
of
L
L
M
s
i
n
h
a
r
dw
a
r
e
de
s
i
gn,”
J
our
nal
of
E
ngi
ne
e
r
i
ng
R
e
s
e
ar
c
h
,
vol
. 13, no. 3, pp.
2392
-
2404
,
2024, doi
:
10.1016/
j
.j
e
r
.2024.08.001.
[
18]
N
. F
l
or
i
a
n, D
. P
ope
s
c
u, a
nd A
. H
o
s
s
u,
“
R
e
a
l
-
t
i
m
e
t
i
r
e
dne
s
s
de
t
e
c
t
i
on s
y
s
t
e
m
us
i
ng
N
V
I
D
I
A
J
e
t
s
on
N
a
no a
nd
O
pe
n
CV
,
”
P
r
oc
e
di
a
C
om
put
e
r
Sc
i
e
n
c
e
, vol
. 242, pp. 536
–
543, 2024, doi
:
10.1016/
j
.pr
oc
s
.2024.08.101.
[
19]
S
.
M
i
t
t
a
l
,
“
A
s
ur
ve
y
on
opt
i
m
i
z
e
d
i
m
pl
e
m
e
nt
a
t
i
on
of
d
e
e
p
l
e
a
r
ni
ng
m
ode
l
s
on
t
he
N
V
I
D
I
A
J
e
t
s
on
pl
a
t
f
or
m
,”
J
our
nal
of
Sy
s
t
e
m
s
A
r
c
hi
t
e
c
t
ur
e
, vol
. 97, pp. 428
–
442, 2019, doi
:
10.1016/
j
.s
ys
a
r
c
.2019.01.011.
[
20]
Y
.
H
u,
H
.
K
i
m
,
K
.
Y
e
,
a
nd
N
.
L
u,
“
A
ppl
yi
ng
f
i
ne
-
t
une
d
LLM
s
f
or
r
e
duc
i
ng
da
t
a
ne
e
ds
i
n
l
oa
d
pr
of
i
l
e
a
na
l
ys
i
s
,”
A
ppl
i
e
d
E
ne
r
gy
,
vol
. 377, 2025, doi
:
10.1016/
j
.a
pe
ne
r
gy.2024.124666.
[
21]
N
.
T
a
yl
or
e
t
al
.
,
“
E
f
f
i
c
i
e
nc
y
a
t
s
c
a
l
e
:
i
nve
s
t
i
ga
t
i
ng
t
he
pe
r
f
or
m
a
nc
e
of
di
m
i
nut
i
ve
l
a
ngua
ge
m
ode
l
s
i
n
c
l
i
ni
c
a
l
t
a
s
ks
,
”
A
r
t
i
f
i
c
i
al
I
nt
e
l
l
i
ge
nc
e
i
n M
e
di
c
i
ne
, vol
. 157, 2024, doi
:
10.1016/
j
.a
r
t
m
e
d.2024.103002.
[
22]
A
.
B
a
e
vs
ki
,
W
.
N
.
H
s
u,
A
.
C
onne
a
u,
a
nd
M
.
A
ul
i
,
“
U
ns
upe
r
vi
s
e
d
s
p
e
e
c
h
r
e
c
ogni
t
i
on,”
A
dv
anc
e
s
i
n
N
e
ur
al
I
nf
or
m
at
i
on
P
r
oc
e
s
s
i
ng Sy
s
t
e
m
s
, vol
. 33, pp. 27826
–
27839, 2021.
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
. 14,
N
o. 5, Oc
to
be
r
2025
:
4202
-
4210
4210
[
23]
Y
.
Z
ha
ng
e
t
al
.
,
“
B
i
gS
S
L
:
e
xpl
or
i
ng
t
he
F
r
ont
i
e
r
of
l
a
r
ge
-
s
c
a
l
e
s
e
m
i
-
s
upe
r
vi
s
e
d
l
e
a
r
ni
ng
f
or
a
ut
om
a
t
i
c
s
pe
e
c
h
r
e
c
ogni
t
i
on,”
I
E
E
E
J
our
nal
on Se
l
e
c
t
e
d T
opi
c
s
i
n Si
gnal
P
r
oc
e
s
s
i
ng
, vol
. 16, no. 6, pp. 1519
–
1532,
2022, doi
:
10.1109/
J
S
T
S
P
.2022.3182537.
[
24]
O
pe
nA
I
, “
I
nt
r
oduc
i
ng w
hi
s
pe
r
,”
O
pe
nA
I
. A
c
c
e
s
s
e
d:
A
ug. 06, 2025. [
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
ope
na
i
.c
om
/
i
nde
x/
w
hi
s
pe
r
/
[
25]
M
.
A
bdi
n
e
t
al
.
,
“
P
hi
-
3
t
e
c
hni
c
a
l
r
e
por
t
:
a
hi
ghl
y
c
a
pa
bl
e
l
a
ngua
ge
m
od
e
l
l
oc
a
l
l
y
on
your
phone
,”
a
r
X
i
v
-
C
om
put
e
r
Sc
i
e
n
c
e
,
pp. 1
-
24, A
ug.
2024
.
[
26]
I
ndr
i
a
ni
,
M
.
H
a
r
r
i
s
,
a
nd
A
.
S
.
A
goe
s
,
“
A
ppl
yi
ng
ha
nd
ge
s
t
ur
e
r
e
c
ogni
t
i
o
n
f
or
us
e
r
gui
de
a
ppl
i
c
a
t
i
on
us
i
ng
m
e
di
a
pi
pe
,”
P
r
oc
e
e
di
ngs
of
t
he
2nd
I
nt
e
r
nat
i
onal
Se
m
i
nar
of
Sc
i
e
nc
e
and
A
ppl
i
e
d
T
e
c
hnol
ogy
(
I
SSA
T
2021)
,
vol
.
207,
2021,
doi
:
10.2991/
a
e
r
.k.211106.017.
[
27]
Y
.
K
or
t
l
i
,
S
.
G
a
bs
i
,
L
.
F
.
C
.
L
.
Y
.
V
oon,
M
.
J
r
i
di
,
M
.
M
e
r
z
ougui
,
a
nd
M
.
A
t
r
i
, “
D
e
e
p
e
m
be
dde
d
hybr
i
d
C
N
N
–
L
S
T
M
ne
t
w
or
k
f
or
l
a
ne
de
t
e
c
t
i
on on
N
V
I
D
I
A
J
e
t
s
on
X
a
vi
e
r
N
X
,”
K
no
w
l
e
dge
-
B
as
e
d Sy
s
t
e
m
s
, vol
.
240, 2022, doi
:
10.1016/
j
.knos
ys
.2021.107941.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Marco
Antonio
Jinete
is
an
Electronic
Engineer
graduated
from
Universidad
Santo
Tomás
in
2007.
He
obtained
a
Master
’
s
degree
in
Engineering
with
a
specialization
in
Industrial
Automatio
n
from
Universidad
Nacional
de
Colombi
a
in
2
014.
He
has
extensive
experience
as
a
project
manager,
researcher,
and
educator,
with
a
stro
ng
focus
on
continu
ous
learning,
investigation,
and
developmen
t.
His
resear
ch
interests
includ
e
image
processing
and
artificial
intell
igence
applied
to
robotic
systems
and
automati
on.
Cu
rrently,
he
works
as
a
professor
and
researcher
,
leading
the
OpenCreator
research
group
focused
on
innovative
developments
in
the
field
of
image
processing.
He
can
be
contacted
at
email:
maajigo@gmail.com
.
Robinson
Jiménez
-
Moreno
is
an
Electronic
Engineer
gra
duated
from
Universidad
Distrital
Francisco
José
de
Caldas
in
2002.
He
received
a
M.Sc.
in
Engineering
from
Universidad
Nacional
de
Colombia
in
2012
and
Ph.D.
in
Engi
neering
at
Universidad
Distrital
Francisco
José
de
Caldas
in
2018.
His
current
working
as
Associate
Professor
of
Universidad
Militar
Nueva
Granada
and
research
focuses
on
the
use
of
convolutional
neural
networks for object rec
ognition and image processing
for robotic appli
cations such as human
-
machine inte
raction. H
e can be
contacte
d at email:
robinson.j
im
enez@
unimilitar.edu.co
.
Anny Astrid
Espitia
-
Cubillos
performed her under
graduate
studie
s in Industrial
Engineering
in
the
Universidad
Militar
Nueva
Granada
in
2002
a
nd
M.Sc.
in
Industrial
Engineering
from
the
Universidad
de
Los
Andes
in
2006.
She
is
an
Associate
Professor
on
Industrial
Engineering
Program
at
Universidad
Militar
Nueva
Granada,
Bogotá,
Colombi
a.
She ca
n be c
ontact
ed at
email:
anny.espi
tia@
unimil
itar.edu.
co
.
Evaluation Warning : The document was created with Spire.PDF for Python.