[PDF]

NeurIPS 2025: A Guide to Key Papers, Trends & Stats PDF Free Download

1 / 16
0 views•16 pages

NeurIPS 2025: A Guide to Key Papers, Trends & Stats PDF Free Download

NeurIPS 2025: A Guide to Key Papers, Trends & Stats PDF free Download. Think more deeply and widely.

N
e
ur
IPS
2025:
A G
u
ide
to
K
e
y
P
a
p
e
rs
,
T
r
e
n
d
s
&
S
t
a
ts
B
y
A
d
r
ie
n
L
a
ur
e
nt
,
CEO
a
t
I
ntu
i
t
i
on
L
ab
s
•
12/2/2025
•
35
m
i
n
r
ead
n
e
ur
i
ps
2025
m
achi
n
e
l
ea
rn
i
n
g ai
c
on
fe
r
e
n
ce
l
a
r
ge
l
a
n
g
u
age
mo
de
ls
ai
r
e
s
ea
r
ch
tr
e
n
d
s
r
ei
n
f
or
ce
m
e
nt
l
ea
rn
i
n
g ai
e
t
hic
s
da
t
a
s
e
ts
a
n
d
be
n
ch
m
a
r
k
s t
e
st
-
o
f
-
t
i
m
e
a
w
a
r
d
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 1 of 16
E
x
ec
ut
i
v
e
S
umm
a
ry
N
e
ur
IPS
(
N
e
ur
a
l
I
n
f
orm
a
t
i
on
P
ro
ce
ss
i
n
g
S
yst
e
ms
)
2025
–
t
he
39
t
h
a
nnu
a
l
m
ee
t
i
n
g
–
r
e
pr
e
s
e
nts
b
ot
h
c
ont
i
nu
i
ty
a
n
d
cha
n
ge
i
n
t
he
worl
d
ʼ
s
pr
e
m
ie
r
m
achi
n
e
l
ea
rn
i
n
g
c
on
fe
r
e
n
ce
.
H
e
l
d
D
ece
m
be
r
2–7,
2025
i
n
S
a
n
D
ieg
o
(
w
i
t
h
a
s
i
mult
a
n
e
ous
s
ec
on
da
ry
s
i
t
e
i
n
M
e
x
ic
o
C
i
ty
),
N
e
ur
IPS
2025
a
ttr
ac
t
ed
ov
e
r
5,200
acce
pt
ed
p
a
p
e
rs
(24.5
%
o
f
~21,575
su
b
m
i
ss
i
ons
)
a
n
d
e
normous
a
tt
e
nt
i
on
f
rom
i
n
d
ustry
,
acade
m
ia
,
a
n
d
m
edia
.
M
aj
or
tr
e
n
d
s
i
n
2025
i
n
c
lu
ded
a
n
e
mp
ha
s
i
s
on
l
a
r
ge
l
a
n
g
u
age
a
n
d
f
oun
da
t
i
on
mo
de
ls
,
r
e
pro
d
u
cibi
l
i
ty
a
n
d
da
t
a
-
ce
ntr
ic
r
e
s
ea
r
ch
v
ia
t
he
g
row
i
n
g
D
a
t
a
s
e
ts
&
B
e
n
chma
r
k
s
T
r
ack
,
a
n
d
e
xpl
ici
t
a
tt
e
nt
i
on
to
so
cie
t
a
l
i
mp
ac
ts
o
f
AI
.
F
or
t
he
fi
rst
t
i
m
e
,
N
e
ur
IPS
i
ntro
d
u
ced
a
P
os
i
t
i
on
P
a
p
e
r
T
r
ack
to
h
ost
p
e
rsp
ec
t
i
v
e
p
iece
s
on
b
ro
ad
AI
i
mpl
ica
t
i
ons
,
a
n
d
a
J
ourn
al
T
r
ack
fea
tur
i
n
g
34
top
p
a
p
e
rs
f
rom
l
eadi
n
g
j
ourn
a
ls
(14
f
rom
JMLR
a
n
d
20
f
rom
A
o
S
)
e
m
bedded
a
s
post
e
rs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
c
on
fe
r
e
n
ce
fea
tur
ed
s
i
x
high
-
pro
fi
l
e
i
nv
i
t
ed
t
a
l
k
s
(
e
.
g
.
b
y
S
utton
,
T
u
fekci
,
C
h
o
i
,
M
i
t
che
ll
,
C
h
o
,
S
a
x
e
)
pro
bi
n
g
f
un
da
m
e
nt
a
l
a
n
d
a
ppl
ied
i
ssu
e
s
i
n
AI
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
M
aj
or
a
w
a
r
d
s
r
ec
o
g
n
i
z
ed
f
ront
ie
r
r
e
s
ea
r
ch
:
t
he
B
e
st
P
a
p
e
r
A
w
a
r
d
w
e
nt
to
Q
i
u
e
t
a
l
.ʼ
s
s
ca
l
ed
LLM
ga
t
i
n
g
wor
k
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
a
n
d
runn
e
r
-
ups
i
n
c
lu
ded
a
n
i
mport
a
nt
a
n
a
lys
i
s
o
f
r
ei
n
f
or
ce
m
e
nt
l
ea
rn
i
n
g
f
or
LLM
r
ea
son
i
n
g
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
I
n
t
he
T
e
st
-
o
f
-
T
ime
ca
t
eg
ory
(
h
onor
i
n
g
2015
p
a
p
e
rs
),
t
he
c
l
a
ss
ic
F
a
st
e
r
R
-
CNN
o
bjec
t
de
t
ec
tor
b
y
R
e
n
e
t
a
l
.
w
a
s
ce
l
eb
r
a
t
ed
,
w
i
t
h
ov
e
r
56,700
ci
t
a
t
i
ons
not
ed
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
N
e
ur
IPS
2025
a
lso
c
ont
i
nu
ed
to
g
r
a
ppl
e
w
i
t
h
pr
ac
t
ica
l
cha
ll
e
n
ge
s
o
f
s
ca
l
e
a
n
d
i
nt
eg
r
i
ty
.
T
he
pro
g
r
a
m
c
omm
i
tt
ee
pro
ce
ss
ed
a
r
ec
or
d
~21.6
k
su
b
m
i
ss
i
ons
w
i
t
h
~20,500
r
e
v
ie
w
e
rs
,
1,663
a
r
ea
chai
rs
,
a
n
d
199
s
e
n
i
or
chai
rs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
o
e
nsur
e
qu
a
l
i
ty
,
chai
rs
i
ntro
d
u
ced
“
R
e
spons
ib
l
e
R
e
v
ie
w
i
n
g
”
pol
icie
s
(
e
.
g
.
p
e
n
a
l
i
z
i
n
g
n
eg
l
ige
nt
r
e
v
ie
w
e
rs
)
a
n
d
r
efi
n
ed
t
hei
r
ca
l
ib
r
a
t
i
on
pro
ce
ss
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
N
ot
ab
ly
,
t
he
c
on
fe
r
e
n
ce
pu
b
l
ic
ly
r
eaffi
rm
ed
str
ic
t
c
on
fide
nt
ia
l
i
ty
af
t
e
r
a
n
O
p
e
n
R
e
v
ie
w
l
eak
,
w
a
rn
i
n
g
t
ha
t
a
ny
e
xplo
i
t
a
t
i
on
o
f
r
e
v
ie
w
e
r
ide
nt
i
t
ie
s
woul
d
be
pun
i
s
hed
s
e
v
e
r
e
ly
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
o
e
m
b
r
ace
n
e
w
a
ut
h
or
i
n
g
tools
,
N
e
ur
IPS
i
ssu
ed
a
de
t
ai
l
ed
LLM
us
age
pol
ic
y
,
a
llow
i
n
g
AI
-
a
ss
i
st
ed
wr
i
t
i
n
g
b
ut
r
e
qu
i
r
i
n
g
a
ut
h
ors
to
v
e
r
if
y
a
n
d
ci
t
e
a
ll
c
ont
e
nt
(
n
e
ur
i
ps
.
cc
).
B
e
yon
d
t
ech
n
ica
l
c
ont
e
nt
,
N
e
ur
IPS
2025
un
de
rs
c
or
ed
c
ommun
i
ty
a
n
d
so
cie
t
a
l
di
m
e
ns
i
ons
.
I
t
h
ost
ed
num
e
rous
affi
n
i
ty
e
v
e
nts
(
e
.
g
.
W
om
e
n
i
n
ML
,
L
a
t
i
n
X
i
n
AI
,
Q
u
ee
r
i
n
AI
,
M
usl
i
ms
i
n
ML
,
e
t
c
.)
to
f
ost
e
r
i
n
c
lus
i
on
(
n
e
ur
i
ps
.
cc
)
(
n
e
ur
i
ps
.
cc
),
a
n
d
8
“
so
cia
l
”
s
e
ss
i
ons
on
top
ic
s
l
ike
AI
s
afe
ty
,
da
t
a
op
e
nn
e
ss
,
or
AI
a
rt
.
I
n
d
ustry
i
nvolv
e
m
e
nt
w
a
s
i
mm
e
ns
e
:
G
oo
g
l
e
a
lon
e
r
e
port
ed
175
acce
pt
ed
p
a
p
e
rs
b
y
i
ts
r
e
s
ea
r
che
rs
a
n
d
sponsor
ed
wor
k
s
h
ops
(
e
.
g
.
W
om
e
n
i
n
ML
,
L
a
t
i
n
X
)
(
r
e
s
ea
r
ch
.
g
oo
g
l
e
);
M
ic
roso
f
t
l
ike
w
i
s
e
st
affed
b
oot
h
s
a
n
d
t
a
l
k
s
;
a
n
d
m
a
ny
c
omp
a
n
ie
s
r
ec
ru
i
t
ed
agg
r
e
ss
i
v
e
ly
(
a
s
not
ed
b
y
pr
e
ss
).
I
n
sum
,
N
e
ur
IPS
2025
w
a
s
a
l
a
n
d
m
a
r
k
e
v
e
nt
,
r
ef
l
ec
t
i
n
g
t
he
m
a
tur
a
t
i
on
o
f
m
achi
n
e
l
ea
rn
i
n
g
:
a
c
on
fe
r
e
n
ce
a
t
g
lo
ba
l
“
m
ega
-
e
v
e
nt
”
s
ca
l
e
,
dee
ply
i
nt
eg
r
a
t
ed
w
i
t
h
i
n
d
ustry
a
n
d
so
cie
t
a
l
c
on
ce
rns
,
y
e
t
st
i
ll
g
roun
ded
i
n
r
ig
orous
s
cie
nt
ific
st
a
n
da
r
d
s
.
T
hi
s
r
e
port
o
ffe
rs
a
n
i
n
-
de
pt
h
a
n
a
lys
i
s
o
f
N
e
ur
IPS
2025:
i
ts
hi
story
a
n
d
or
ga
n
i
z
a
t
i
on
;
su
b
m
i
ss
i
on
a
n
d
r
e
v
ie
w
pro
ce
ss
e
s
;
ke
y
t
he
m
e
s
a
n
d
fi
n
di
n
g
s
f
rom
t
he
t
ech
n
ica
l
pro
g
r
a
m
;
ca
s
e
stu
die
s
(
be
st
p
a
p
e
rs
,
i
nv
i
t
ed
t
a
l
k
s
,
wor
k
s
h
ops
);
a
n
d
di
s
c
uss
i
on
o
f
i
mpl
ica
t
i
ons
f
or
t
he
f
utur
e
o
f
AI
r
e
s
ea
r
ch
.
A
ll
o
b
s
e
rv
a
t
i
ons
a
r
e
support
ed
b
y
da
t
a
f
rom
t
he
o
fficia
l
N
e
ur
IPS
pu
b
l
ica
t
i
ons
,
a
ut
h
or
i
t
a
t
i
v
e
n
e
ws
,
a
n
d
e
xp
e
rt
r
e
ports
.
1.
I
ntro
d
u
c
t
i
on
a
n
d
B
ackg
roun
d
T
he
C
on
fe
r
e
n
ce
on
N
e
ur
al
I
n
f
or
ma
t
i
on
P
ro
ce
ss
i
n
g
S
yst
em
s
(
N
e
ur
IPS
,
f
orm
e
rly
NIP
S
)
ha
s
bee
n
he
l
d
a
nnu
a
lly
s
i
n
ce
1987
a
n
d
i
s
w
ide
ly
r
ec
o
g
n
i
z
ed
a
s
t
he
pr
e
m
ie
r
acade
m
ic
c
on
fe
r
e
n
ce
i
n
m
achi
n
e
l
ea
rn
i
n
g
a
n
d
c
omput
a
t
i
on
a
l
n
e
uros
cie
n
ce
(
st
a
t
ic
.
h
lt
.
b
m
e
.
h
u
).
F
oun
ded
a
s
a
n
i
nt
e
r
di
s
ci
pl
i
n
a
ry
m
ee
t
i
n
g
on
n
e
ur
a
l
n
e
twor
k
s
,
N
e
ur
IPS
ha
s
g
rown
e
xpon
e
nt
ia
lly
ov
e
r
t
he
decade
s
:
b
y
2016
i
t
h
ost
ed
~5,000
p
a
rt
ici
p
a
nts
,
b
y
2017
~8,000
(
st
a
t
ic
.
h
lt
.
b
m
e
.
h
u
),
a
n
d
i
ts
v
e
nu
e
s
ha
v
e
e
xp
a
n
ded
be
yon
d
s
ki
r
e
sorts
to
m
aj
or
c
onv
e
nt
i
on
ce
nt
e
rs
.
I
n
r
ece
nt
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 2 of 16
y
ea
rs
,
N
e
ur
IPS
ha
s
bec
om
e
a
ba
rom
e
t
e
r
f
or
t
he
AI
fie
l
d
ʼ
s
g
rowt
h
a
n
d
cha
ll
e
n
ge
s
.
F
or
e
x
a
mpl
e
,
N
e
ur
IPS
2023
i
n
N
e
w
O
rl
ea
ns
d
r
e
w
ov
e
r
16
,
000
a
tt
e
n
dee
s
(
[1]
f
ortun
e
.
c
om
)
(
f
rom
rou
gh
ly
13,330
p
a
p
e
r
su
b
m
i
ss
i
ons
,
up
38
%
f
rom
2022
(
[1]
f
ortun
e
.
c
om
))
–
fa
r
be
yon
d
i
ts
f
oun
de
rs
ʼ
e
xp
ec
t
a
t
i
ons
.
A R
e
ut
e
rs
r
e
port
de
s
c
r
ibed
N
e
ur
IPS
2024
a
s
“
t
he
l
a
r
ge
st
a
nnu
a
l
ga
t
he
r
i
n
g
f
or
AI
”
(
[2]
www
.
r
e
ut
e
rs
.
c
om
).
T
hi
s
r
a
p
id
e
xp
a
ns
i
on
r
ef
l
ec
ts
two
d
r
i
v
e
rs
:
(1)
a
sur
ge
o
f
i
nt
e
r
e
st
a
n
d
i
nv
e
stm
e
nt
i
n
AI
(
e
sp
ecia
lly
d
r
i
v
e
n
b
y
ge
n
e
r
a
t
i
v
e
mo
de
ls
a
n
d
i
n
d
ustry
i
nvolv
e
m
e
nt
),
a
n
d
(2)
t
he
b
ro
ade
n
i
n
g
s
c
op
e
o
f
N
e
ur
IPS
i
ts
e
l
f
(
w
i
t
h
n
e
w
tr
ack
s
,
wor
k
s
h
ops
,
a
n
d
c
ommun
i
ty
pro
g
r
a
ms
).
T
he
pro
g
r
a
m
now
i
n
c
lu
de
s
mult
i
pl
e
p
a
r
a
ll
e
l
tr
ack
s
(
e
.
g
.
m
ai
n
r
e
s
ea
r
ch
,
da
t
a
s
e
ts
&
be
n
ch
m
a
r
k
s
,
pos
i
t
i
on
p
a
p
e
rs
),
tutor
ia
ls
,
wor
k
s
h
ops
,
a
n
d
e
v
e
n
a
j
ourn
al
tr
ack
b
r
idgi
n
g
ML
a
n
d
st
a
t
i
st
ic
s
j
ourn
a
ls
.
T
he
c
on
fe
r
e
n
ce
ʼ
s
f
orm
a
t
ha
s
e
volv
ed
:
st
i
ll
root
ed
i
n
p
ee
r
-
r
e
v
ie
w
ed
p
a
p
e
r
pr
e
s
e
nt
a
t
i
ons
,
i
t
now
a
lso
fea
tur
e
s
a
v
ib
r
a
nt
e
xpo
,
so
cia
l
e
v
e
nts
,
a
n
d
pro
fe
ss
i
on
a
l
d
r
if
ts
.
A
s
o
f
D
ece
m
be
r
2025,
N
e
ur
IPS
2025
ha
s
t
ake
n
pl
ace
,
o
ffe
r
i
n
g
a
c
ompr
ehe
ns
i
v
e
sn
a
ps
h
ot
o
f
t
he
AI
c
ommun
i
ty
ʼ
s
st
a
t
e
.
T
he
c
on
fe
r
e
n
ce
w
a
s
head
qu
a
rt
e
r
ed
a
t
t
he
S
a
n
D
ieg
o
C
onv
e
nt
i
on
C
e
nt
e
r
(
D
ece
m
be
r
2–7,
2025)
w
i
t
h
a
c
o
-
lo
ca
t
ed
s
i
t
e
i
n
M
e
x
ic
o
C
i
ty
(
N
ov
30
–
D
ec
5)
(
n
e
ur
i
ps
.
cc
).
I
t
w
a
s
or
ga
n
i
z
ed
b
y
t
he
N
e
ur
a
l
I
n
f
orm
a
t
i
on
P
ro
ce
ss
i
n
g
S
yst
e
ms
F
oun
da
t
i
on
a
n
d
chai
r
ed
b
y
a
l
a
r
ge
c
omm
i
tt
ee
(
mult
i
pl
e
pro
g
r
a
m
tr
ack
s
each
w
i
t
h
chai
rs
).
T
hi
s
r
e
port
prov
ide
s
a
n
e
xt
e
ns
i
v
e
a
n
a
lys
i
s
o
f
N
e
ur
IPS
2025,
c
ov
e
r
i
n
g
all
k
nown
a
sp
ec
ts
:
hi
stor
ica
l
c
ont
e
xt
,
c
on
fe
r
e
n
ce
or
ga
n
i
z
a
t
i
on
,
su
b
m
i
ss
i
on
a
n
d
r
e
v
ie
w
de
t
ai
ls
,
ke
y
p
a
rt
ici
p
a
nts
a
n
d
sponsors
,
pro
g
r
a
m
c
ont
e
nt
(
p
a
p
e
rs
,
t
a
l
k
s
,
wor
k
s
h
ops
),
c
ommun
i
ty
i
n
i
t
ia
t
i
v
e
s
,
a
n
d
f
utur
e
di
r
ec
t
i
ons
.
I
t
i
nt
eg
r
a
t
e
s
da
t
a
f
rom
o
fficia
l
sour
ce
s
(
N
e
ur
IPS
w
eb
s
i
t
e
s
a
n
d
b
lo
g
s
),
s
ec
on
da
ry
n
e
ws
a
n
a
lys
i
s
,
a
n
d
pu
b
l
i
s
hed
p
a
p
e
r
ab
str
ac
ts
.
2.
N
e
ur
IPS
2025
O
r
ga
n
i
z
a
t
i
on
a
n
d
S
c
op
e
D
a
t
e
s
a
n
d
V
e
nu
e
s
.
N
e
ur
IPS
2025
w
a
s
t
he
39
t
h
C
on
fe
r
e
n
ce
a
n
d
too
k
pl
ace
D
ec
.
2–7,
2025
i
n
S
a
n
D
ieg
o
,
CA
(
S
a
n
D
ieg
o
C
onv
e
nt
i
on
C
e
nt
e
r
).
P
a
r
a
ll
e
l
to
t
he
m
ai
n
s
i
t
e
,
a
N
e
ur
IP
S
M
e
x
ic
o
C
i
ty
(
CDM
X
)
2025
e
v
e
nt
w
a
s
he
l
d
N
ov
.
30
–
D
ec
.
5,
2025
a
t
H
i
lton
M
e
x
ic
o
C
i
ty
R
ef
orm
a
(
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
w
a
s
t
he
fi
rst
t
i
m
e
N
e
ur
IPS
he
l
d
a
“
s
ec
on
da
ry
lo
ca
t
i
on
”
outs
ide
i
ts
m
ai
n
U
S
/
C
a
n
ada
s
i
t
e
.
M
e
x
ic
o
C
i
ty
h
ost
ed
a
num
be
r
o
f
wor
k
s
h
ops
a
n
d
tutor
ia
ls
(
e
.
g
.
on
v
i
s
i
on
-
l
a
n
g
u
age
mo
de
ls
,
ro
b
ot
ic
s
,
r
e
spons
ib
l
e
AI
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
s
h
ow
i
n
g
t
he
c
on
fe
r
e
n
ce
ʼ
s
e
xp
a
n
di
n
g
g
lo
ba
l
f
ootpr
i
nt
.
S
ca
l
e
a
n
d
A
tt
e
n
da
n
ce
.
A
lt
h
ou
gh
o
fficia
l
fi
n
a
l
a
tt
e
n
da
n
ce
fig
ur
e
s
f
or
2025
a
r
e
not
y
e
t
pu
b
l
i
s
hed
,
t
he
c
on
fe
r
e
n
ce
sol
d
out
r
egi
str
a
t
i
on
(
n
e
ur
i
ps
.
cc
)
a
n
d
c
ont
i
nu
ed
t
he
upw
a
r
d
tr
e
n
d
o
f
r
ece
nt
y
ea
rs
.
B
y
a
n
a
lo
g
y
,
N
e
ur
IPS
2024
i
n
V
a
n
c
ouv
e
r
d
r
e
w
~16,000
a
tt
e
n
dee
s
(
[2]
www
.
r
e
ut
e
rs
.
c
om
).
T
he
a
ut
h
or
a
n
d
r
e
v
ie
w
e
r
c
ounts
i
llustr
a
t
e
t
he
s
ca
l
e
:
t
he
N
e
ur
IPS
2025
M
ai
n
T
r
ack
r
ecei
v
ed
a
st
agge
r
i
n
g
21
,
575
v
alid
su
bmi
ss
i
ons
,
o
f
w
hich
5
,
290
w
e
r
e
acce
pt
ed
(24.52
%
acce
pt
a
n
ce
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
S
upport
i
n
g
t
hi
s
m
a
ss
i
v
e
r
e
v
ie
w
i
n
g
lo
ad
w
e
r
e
20
,
518
r
e
v
ie
w
e
rs
,
1,663
a
r
ea
chai
rs
(
AC
s
),
a
n
d
199
s
e
n
i
or
a
r
ea
chai
rs
(
SAC
s
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
G
oo
g
l
e
(
a
D
ia
mon
d
S
ponsor
)
not
ed
i
t
had
175
acce
pt
ed
p
a
p
e
rs
sp
a
nn
i
n
g
N
e
ur
IPS
2025ʼ
s
pro
g
r
a
ms
,
plus
i
nvolv
e
m
e
nt
i
n
ov
e
r
70
wor
k
s
h
ops
,
tutor
ia
ls
a
n
d
c
omp
e
t
i
t
i
ons
(
r
e
s
ea
r
ch
.
g
oo
g
l
e
).
A
ll
o
f
t
hi
s
un
de
rs
c
or
e
s
t
he
c
on
fe
r
e
n
ce
ʼ
s
d
u
a
l
n
a
tur
e
:
a
n
acade
m
ic
f
orum
o
f
unpr
ecede
nt
ed
s
i
z
e
,
a
n
d
a
m
aj
or
i
n
d
ustry
s
h
ow
ca
s
e
w
i
t
h
c
orpor
a
t
e
R
&
D
hea
v
i
ly
r
e
pr
e
s
e
nt
ed
.
T
ab
l
e
1
summ
a
r
i
z
e
s
ke
y
st
a
t
i
st
ic
s
o
f
N
e
ur
IPS
2025:
S
t
a
t
i
st
ic
V
a
lu
e
(
N
e
ur
IPS
2025)
M
ai
n
T
r
ack
S
u
b
m
i
ss
i
ons
21,575
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
M
ai
n
T
r
ack
A
cce
pt
ed
(
P
a
p
e
rs
) 5,290
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
M
ai
n
T
r
ack
A
cce
pt
a
n
ce
R
a
t
e
24.5
%
(~21,575
→
5,290)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
T
ot
a
l
R
e
v
ie
w
e
rs
(
a
ll
tr
ack
s
) 20,518
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
A
r
ea
C
hai
rs
/
S
e
n
i
or
A
r
ea
C
hai
rs
1,663
AC
s
;
199
SAC
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 3 of 16
S
t
a
t
i
st
ic
V
a
lu
e
(
N
e
ur
IPS
2025)
D
a
t
a
s
e
ts
&
B
e
n
ch
m
a
r
k
s
T
r
ack
S
u
b
m
i
ss
i
ons
1,995
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
P
os
i
t
i
on
P
a
p
e
r
T
r
ack
S
u
b
m
i
ss
i
ons
(
n
e
w
;
s
ee
S
ec
t
i
on
4)
A
ffi
n
i
ty
E
v
e
nts
7
e
v
e
nts
(
e
.
g
.
W
om
e
n
i
n
ML
,
Q
u
ee
r
i
n
AI
)
(
n
e
ur
i
ps
.
cc
)
(
n
e
ur
i
ps
.
cc
)
I
nv
i
t
ed
K
e
ynot
e
s
6
sp
eake
rs
(
S
utton
,
T
u
fekci
,
C
h
o
i
,
M
i
t
che
ll
,
C
h
o
,
S
a
x
e
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
T
able
1
:
K
e
y
fac
ts
ab
out
N
e
ur
IP
S
2025
.
S
i
t
e
-
ci
t
ed
st
a
t
i
st
ic
s
c
om
e
f
rom
N
e
ur
IPS
o
fficia
l
c
ommun
ica
t
i
ons
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
a
mon
g
ot
he
rs
.
O
r
ga
n
i
z
i
n
g
C
omm
i
tt
ee
a
n
d
T
r
ack
s
.
T
he
c
on
fe
r
e
n
ce
w
a
s
m
a
n
aged
b
y
a
l
a
r
ge
or
ga
n
i
z
i
n
g
c
omm
i
tt
ee
.
N
a
n
c
y
C
he
n
,
M
a
rzy
eh
G
ha
ss
e
m
i
,
P
i
otr
K
on
i
usz
,
R
a
zv
a
n
P
a
s
ca
nu
,
a
n
d
H
su
a
n
-
T
ie
n
L
i
n
w
e
r
e
pro
g
r
a
m
c
o
-
chai
rs
(
n
e
ur
i
ps
.
cc
),
support
ed
b
y
d
oz
e
ns
o
f
a
r
ea
chai
rs
.
T
he
r
e
w
e
r
e
mult
i
pl
e
p
a
r
a
ll
e
l
“
tr
ack
s
”:
M
ai
n
T
r
ack
(
R
e
s
ea
r
ch
P
a
p
e
rs
):
T
r
adi
t
i
on
a
l
r
efe
r
eed
su
b
m
i
ss
i
ons
o
f
a
l
g
or
i
t
h
m
ic
ML
r
e
s
ea
r
ch
.
D
a
t
a
s
e
ts
&
B
e
n
ch
m
a
r
k
s
T
r
ack
:
C
ont
i
nu
e
s
s
i
n
ce
2021
to
f
o
c
us
on
da
t
a
-
a
n
d
e
v
a
lu
a
t
i
on
-
ce
ntr
ic
c
ontr
ib
ut
i
ons
.
(
T
hi
s
tr
ack
g
r
e
w
to
1,995
su
b
m
i
ss
i
ons
i
n
2025
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
a
n
d
i
ntro
d
u
ced
r
ig
orous
da
t
a
s
ha
r
i
n
g
st
a
n
da
r
d
s
.)
P
os
i
t
i
on
P
a
p
e
r
T
r
ack
(
NE
W
2025):
A
n
e
w
tr
ack
f
or
p
e
rsp
ec
t
i
v
e
/
op
i
n
i
on
p
a
p
e
rs
on
AI
ʼ
s
b
ro
ad
i
mp
ac
ts
;
i
n
i
t
ia
t
ed
i
n
2025
b
y
tr
ack
chai
rs
B
a
r
ba
r
a
E
n
ge
l
ha
r
d
t
a
n
d
C
ynt
hia
R
u
di
n
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
I
t
sol
ici
t
ed
c
r
i
t
ica
l
v
ie
wpo
i
nts
on
l
ega
l
,
e
t
hica
l
,
a
n
d
so
cie
t
a
l
i
ssu
e
s
(
c
opyr
igh
t
,
e
nv
i
ronm
e
nt
a
l
c
ost
,
e
t
c
.)
a
n
d
w
i
ll
ha
v
e
i
ts
acce
pt
ed
p
a
p
e
rs
i
nt
eg
r
a
t
ed
i
nto
t
he
m
ai
n
pro
g
r
a
m
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
J
ourn
a
l
T
r
ack
(
NE
W
2025):
A
nov
e
l
tr
ack
pr
e
s
e
nt
i
n
g
s
e
l
ec
t
ed
outst
a
n
di
n
g
p
a
p
e
rs
r
ece
ntly
pu
b
l
i
s
hed
i
n
m
aj
or
j
ourn
a
ls
.
N
e
ur
IPS
2025
fea
tur
ed
34
p
a
p
e
rs
(14
f
rom
JMLR
,
20
f
rom
A
nn
a
ls
o
f
S
t
a
t
i
st
ic
s
)
a
s
post
e
rs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
str
e
n
g
t
he
n
i
n
g
t
ie
s
be
tw
ee
n
t
he
ML
a
n
d
st
a
t
i
st
ic
s
c
ommun
i
t
ie
s
.
W
or
k
s
h
ops
a
n
d
T
utor
ia
ls
:
O
v
e
r
80
wor
k
s
h
ops
/
tutor
ia
ls
ac
ross
a
ll
da
ys
,
m
a
ny
propos
ed
v
ia
a
c
omp
e
t
i
t
i
v
e
pro
ce
ss
.
T
op
ic
s
r
a
n
ged
f
rom
e
m
b
o
died
AI
,
V
LM
s
,
to
so
cia
lly
r
e
spons
ib
l
e
AI
.
N
ot
ab
ly
,
wor
k
s
h
ops
/
tutor
ia
ls
w
e
r
e
c
o
-
lo
ca
t
ed
i
n
b
ot
h
S
a
n
D
ieg
o
a
n
d
M
e
x
ic
o
C
i
ty
(
s
ee
S
ec
t
i
on
6).
S
e
p
a
r
a
t
e
f
rom
t
he
t
ech
n
ica
l
tr
ack
s
w
e
r
e
affi
n
i
ty
a
n
d
so
cial
e
v
e
nts
,
de
s
ig
n
ed
to
b
u
i
l
d
c
ommun
i
ty
(
s
ee
S
ec
t
i
on
7).
A
lso
,
E
x
hibi
tors
/
E
xpo
–
a
c
omm
e
r
cia
l
e
x
hibi
t
i
on
w
he
r
e
sponsors
de
monstr
a
t
ed
pro
d
u
c
ts
a
n
d
r
ec
ru
i
t
ed
t
a
l
e
nt
–
r
a
n
t
h
rou
gh
out
t
he
w
eek
.
T
he
c
on
fe
r
e
n
ce
a
pp
a
n
d
pr
i
nt
ed
pro
g
r
a
m
g
u
ided
a
tt
e
n
dee
s
t
h
rou
gh
h
un
d
r
ed
s
o
f
s
e
ss
i
ons
.
C
omp
a
r
i
son
to
P
a
st
Y
ea
rs
.
T
he
g
rowt
h
f
rom
pr
e
v
i
ous
y
ea
rs
i
s
st
a
r
k
.
S
u
b
m
i
ss
i
ons
j
ump
ed
f
rom
9,467
i
n
2020
to
21,575
i
n
2025
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
s
F
ortun
e
m
aga
z
i
n
e
not
ed
,
N
e
ur
IPS
w
a
s
on
ce
“
t
he
g
o
-
to
f
or
l
a
t
e
st
r
e
s
ea
r
ch
,”
b
ut
post
-2020
ha
s
bec
om
e
a
v
a
st
r
ec
ru
i
t
i
n
g
a
n
d
i
n
d
ustr
ia
l
e
xtr
a
v
aga
nz
a
(
[1]
f
ortun
e
.
c
om
).
I
n
deed
,
m
a
ny
2023/24
n
e
ws
a
rt
ic
l
e
s
e
mp
ha
s
i
z
e
big
t
ech
a
n
d
fi
n
a
n
ce
p
a
rt
ici
p
a
t
i
on
.
B
y
2025,
N
e
ur
IPS
c
l
ea
rly
f
un
c
t
i
ons
a
s
b
ot
h
a
c
utt
i
n
g
-
edge
r
e
s
ea
r
ch
v
e
nu
e
a
n
d
a
g
lo
ba
l
i
n
d
ustry
summ
i
t
f
or
AI
e
n
gi
n
ee
r
i
n
g
,
t
a
l
e
nt
,
a
n
d
pol
ic
y
.
I
n
summ
a
ry
,
N
e
ur
IPS
2025
w
a
s
a
spr
a
wl
i
n
g
e
v
e
nt
:
t
h
ous
a
n
d
s
o
f
p
a
rt
ici
p
a
nts
f
rom
acade
m
ia
a
n
d
i
n
d
ustry
,
d
oz
e
ns
o
f
p
a
r
a
ll
e
l
tr
ack
s
,
a
n
d
s
c
or
e
s
o
f
so
cia
l
pro
g
r
a
ms
.
T
he
r
e
m
ai
n
de
r
o
f
t
hi
s
r
e
port
e
x
a
m
i
n
e
s
t
he
de
t
ai
ls
o
f
t
he
s
e
c
ompon
e
nts
a
n
d
t
hei
r
i
mpl
ica
t
i
ons
.
3.
S
u
b
m
i
ss
i
on
a
n
d
R
e
v
ie
w
P
ro
ce
ss
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 4 of 16
3.1
S
u
b
m
i
ss
i
on
S
t
a
t
i
st
ic
s
T
he
N
e
ur
IPS
2025
ca
ll
f
or
p
a
p
e
rs
f
ollow
ed
t
he
tr
adi
t
i
on
a
l
s
ched
ul
e
(
ab
str
ac
t
d
u
e
M
a
y
11,
2025;
f
ull
p
a
p
e
rs
M
a
y
15,
2025).
I
n
r
e
spons
e
,
t
he
M
ai
n
T
r
ack
r
ecei
v
ed
21,575
v
a
l
id
p
a
p
e
r
su
b
m
i
ss
i
ons
b
y
t
he
dead
l
i
n
e
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
a
rou
gh
ly
61
%
i
n
c
r
ea
s
e
ov
e
r
2024.
S
u
ch
unpr
ecede
nt
ed
volum
e
str
ai
n
ed
r
e
sour
ce
s
b
ut
r
ef
l
ec
ts
t
he
g
lo
ba
l
sur
ge
i
n
AI
r
e
s
ea
r
ch
pro
d
u
c
t
i
v
i
ty
.
A
f
t
e
r
p
ee
r
r
e
v
ie
w
a
n
d
ca
l
ib
r
a
t
i
on
,
5,290
p
a
p
e
rs
w
e
r
e
acce
pt
ed
,
y
ie
l
di
n
g
a
n
acce
pt
a
n
ce
r
a
t
e
o
f
~24.52
%
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
N
ot
ab
ly
,
t
hi
s
acce
pt
a
n
ce
r
a
t
e
i
s
c
omp
a
r
ab
l
e
to
pr
e
v
i
ous
y
ea
rs
(~24–25
%
)
de
sp
i
t
e
t
he
d
ou
b
l
ed
su
b
m
i
ss
i
ons
,
i
n
dica
t
i
n
g
t
he
pro
g
r
a
m
ai
m
ed
to
e
xp
a
n
d
ca
p
aci
ty
r
a
t
he
r
t
ha
n
t
igh
t
e
n
s
e
l
ec
t
i
v
i
ty
.
(
P
ro
g
r
a
m
chai
rs
e
xpl
ici
tly
not
ed
t
he
y
c
oul
d
ha
v
e
acce
pt
ed
mor
e
if
n
eeded
;
t
he
v
e
nu
e
s
i
z
e
did
not
c
onstr
ai
n
acce
pt
a
n
ce
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).)
R
ef
l
ec
t
i
n
g
t
hi
s
s
ca
l
e
,
N
e
ur
IPS
a
ss
e
m
b
l
ed
a
n
unpr
ecede
nt
ed
r
e
v
ie
w
c
orps
.
T
he
M
ai
n
T
r
ack
had
20
,
518
r
e
v
ie
w
e
rs
,
1
,
663
a
r
ea
chai
rs
,
a
n
d
199
s
e
n
i
or
a
r
ea
chai
rs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
M
u
ch
o
f
t
hi
s
had
to
be
r
ec
ru
i
t
ed
sw
if
tly
:
h
un
d
r
ed
s
o
f
n
e
w
r
e
v
ie
w
e
rs
a
n
d
fi
rst
-
t
i
m
e
AC
s
j
o
i
n
ed
to
m
a
n
age
t
he
lo
ad
.
T
he
da
t
a
s
e
ts
&
be
n
ch
m
a
r
k
s
tr
ack
s
i
m
i
l
a
rly
s
a
w
r
a
p
id
g
rowt
h
–
i
t
had
ab
out
2
,
680
r
e
v
ie
w
e
rs
support
i
n
g
1
,
995
su
bmi
ss
i
ons
i
n
2025
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
ll
i
n
a
ll
,
t
e
ns
o
f
t
h
ous
a
n
d
s
o
f
c
ommun
i
ty
m
e
m
be
rs
c
ontr
ib
ut
ed
to
r
e
v
ie
ws
.
A
cce
pt
a
n
ce
R
a
t
e
s
b
y
T
r
ack
.
T
he
ov
e
r
a
ll
acce
pt
a
n
ce
(5,290/21,575)
~24.5
%
w
a
s
s
i
m
i
l
a
r
to
pr
e
v
i
ous
N
e
ur
IPS
c
y
c
l
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
(
F
or
r
efe
r
e
n
ce
,
F
ortun
e
r
e
port
ed
2023
a
t
~13,330
su
b
s
w
i
t
h
3,753
acce
pt
ed
[~28
%
]
(
[1]
f
ortun
e
.
c
om
).)
T
he
D
a
t
a
s
e
ts
T
r
ack
hi
stor
ica
lly
ha
s
highe
r
acce
pt
a
n
ce
(
r
ef
l
ec
t
i
n
g
on
e
-
s
ided
r
e
v
ie
w
,
d
ou
b
l
e
-
b
l
i
n
d
,
e
t
c
.),
b
ut
e
x
ac
t
2025
fig
ur
e
s
w
e
r
e
not
di
s
c
los
ed
pu
b
l
ic
ly
.
T
he
n
e
w
P
os
i
t
i
on
P
a
p
e
r
T
r
ack
l
ike
ly
had
a
high
pr
e
l
i
m
i
n
a
ry
acce
pt
a
n
ce
to
e
nsur
e
su
fficie
nt
c
ont
e
nt
;
i
ts
CFP
su
gge
st
ed
a
s
i
n
g
l
e
-
b
l
i
n
d
r
e
v
ie
w
m
echa
n
i
sm
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
S
u
bjec
t
A
r
ea
s
.
W
hi
l
e
N
e
ur
IPS
su
b
m
i
ss
i
on
ca
t
eg
or
ie
s
a
r
e
b
ro
ad
(
e
.
g
.
l
ea
rn
i
n
g
t
he
ory
,
sp
eech
,
ro
b
ot
ic
s
,
v
i
s
i
on
,
e
t
c
.),
t
he
2025
c
y
c
l
e
s
h
ow
ed
a
pronoun
ced
t
i
lt
tow
a
r
d
ge
n
e
r
a
t
i
v
e
a
n
d
f
oun
da
t
i
on
m
o
del
s
,
l
a
r
ge
-
s
ca
l
e
m
achi
n
e
l
ea
rn
i
n
g
,
a
n
d
i
nt
e
r
di
s
ci
pl
i
n
a
ry
a
ppl
ica
t
i
ons
.
F
or
e
x
a
mpl
e
,
m
a
ny
acce
pt
ed
p
a
p
e
rs
f
o
c
us
ed
on
l
a
n
g
u
age
mo
de
ls
,
mult
i
mo
da
l
AI
,
r
ei
n
f
or
ce
m
e
nt
l
ea
rn
i
n
g
,
a
n
d
r
ea
l
-
worl
d
de
ploym
e
nt
i
ssu
e
s
.
T
hi
s
a
l
ig
ns
w
i
t
h
i
nv
i
t
ed
t
a
l
k
t
he
m
e
s
(
s
ee
S
ec
t
i
on
5).
T
he
D
a
t
a
s
e
ts
tr
ack
r
e
port
ed
t
ha
t
84
%
o
f
acce
pt
ed
da
t
a
s
e
ts
i
ntro
d
u
ced
n
e
w
su
cce
ssor
be
n
ch
m
a
r
k
s
(
r
a
t
he
r
t
ha
n
r
e
p
ackagi
n
g
ol
d
t
a
s
k
s
),
o
f
t
e
n
l
i
n
ked
to
LLM
e
v
a
lu
a
t
i
on
or
s
cie
n
ce
d
om
ai
ns
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
3.2
R
e
v
ie
w
P
ro
ce
ss
a
n
d
R
e
spons
ib
l
e
P
r
ac
t
ice
s
N
e
ur
IPS
e
mploys
a
mult
i
-
t
ie
r
r
e
v
ie
w
pro
ce
ss
:
a
ss
ig
n
ed
r
e
v
ie
w
e
rs
fi
rst
e
v
a
lu
a
t
e
p
a
p
e
rs
,
w
i
t
h
a
ut
h
ors
su
b
m
i
tt
i
n
g
r
e
spons
e
s
(
r
eb
utt
a
ls
),
a
n
d
AC
s
a
n
d
SAC
s
m
aki
n
g
fi
n
a
l
r
ec
omm
e
n
da
t
i
ons
.
T
he
unpr
ecede
nt
ed
s
ca
l
e
i
ntro
d
u
ced
cha
ll
e
n
ge
s
i
n
c
ons
i
st
e
n
c
y
a
n
d
e
xp
e
rt
i
s
e
.
I
n
a
post
-
h
o
c
r
ef
l
ec
t
i
on
,
t
he
P
ro
g
r
a
m
C
omm
i
tt
ee
s
(
PC
)
chai
rs
not
ed
t
ha
t
s
ca
l
i
n
g
up
fie
l
d
s
(
e
sp
ecia
lly
e
m
e
r
ge
nt
on
e
s
l
ike
LLM
s
)
r
i
s
ked
i
n
e
xp
e
r
ie
n
ced
r
e
v
ie
w
e
rs
or
top
ic
mi
s
ma
t
che
s
(
e
.
g
.
a
n
ol
d
-
g
u
a
r
d
r
e
v
ie
w
e
r
i
ll
-
su
i
t
ed
to
a
n
e
w
mult
i
mo
da
l
p
a
p
e
r
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
o
m
a
n
age
t
hi
s
,
2025
i
ntro
d
u
ced
s
e
v
e
r
a
l
n
e
w
r
e
spons
ib
l
e
r
e
v
ie
w
i
n
g
i
n
i
t
ia
t
i
v
e
s
.
A
ut
h
ors
w
i
t
h
de
l
i
nqu
e
nt
r
e
v
ie
w
i
n
g
d
ut
ie
s
c
oul
d
ha
v
e
t
hei
r
own
su
b
m
i
ss
i
ons
p
e
n
a
l
i
z
ed
;
g
rossly
n
eg
l
ige
nt
r
e
v
ie
w
e
rs
(
e
.
g
.
ig
nor
ed
dead
l
i
n
e
s
,
prov
ided
unpro
fe
ss
i
on
a
l
r
e
v
ie
ws
)
woul
d
tr
igge
r
a
utom
a
t
ic
r
ejec
t
i
on
o
f
a
c
o
-
a
ut
h
or
ed
p
a
p
e
r
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
m
i
rrors
pol
icie
s
a
t
ot
he
r
top
c
on
fe
r
e
n
ce
s
.
T
he
chai
rs
a
lso
m
i
n
i
m
i
z
ed
“
biddi
n
g
”
r
e
l
ia
n
ce
to
a
vo
id
c
ollus
i
on
,
a
n
d
us
ed
mor
e
a
utom
a
t
ed
m
a
t
chi
n
g
o
f
r
e
v
ie
w
e
rs
to
p
a
p
e
rs
,
acce
pt
i
n
g
som
e
no
i
s
e
f
or
i
n
c
r
ea
s
ed
s
afe
ty
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
chai
rs
stu
ck
c
los
e
ly
to
a
ca
l
ib
r
a
t
i
on
pro
ce
ss
:
w
he
n
r
e
v
ie
w
e
rs
/
AC
s
c
on
f
l
ic
t
ed
,
SAC
s
a
n
d
PC
s
m
a
nu
a
lly
adj
u
dica
t
ed
b
or
de
rl
i
n
e
ca
s
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
y
trust
ed
e
xp
e
r
ie
n
ced
AC
s
to
figh
t
f
or
stron
g
p
a
p
e
rs
,
b
ut
som
e
t
i
m
e
s
ov
e
rturn
ed
high
-
r
e
v
ie
w
e
r
s
c
or
e
s
w
he
n
AC
s
or
SAC
s
ide
nt
ified
s
e
r
i
ous
f
l
a
ws
.
T
he
chai
rs
e
mp
ha
s
i
z
ed
t
ha
t
w
hi
l
e
som
e
deci
s
i
ons
m
igh
t
di
s
a
ppo
i
nt
a
ut
h
ors
,
t
he
y
ai
m
ed
to
m
ai
nt
ai
n
lon
g
-
e
st
ab
l
i
s
hed
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 5 of 16
N
e
ur
IPS
st
a
n
da
r
d
s
o
f
nov
e
lty
a
n
d
t
ech
n
ica
l
soun
d
n
e
ss
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
y
not
ed
t
ha
t
out
o
f
m
a
ny
c
ont
e
nt
i
ous
ca
s
e
s
,
only
11
p
a
p
e
rs
w
e
r
e
r
e
v
e
rs
ed
f
or
s
e
v
e
r
e
acade
m
ic
m
i
s
c
on
d
u
c
t
b
y
a
c
o
-
a
ut
h
or
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
su
gge
st
i
n
g
t
he
n
e
w
pol
icie
s
w
e
r
e
i
n
deed
us
ed
to
e
n
f
or
ce
i
nt
eg
r
i
ty
.
3.3
E
t
hic
s
,
C
on
fide
nt
ia
l
i
ty
,
a
n
d
LLM
U
s
e
N
e
ur
IPS
2025
r
eaffi
rm
ed
str
ic
t
c
o
de
s
o
f
c
on
d
u
c
t
a
n
d
c
on
fide
nt
ia
l
i
ty
.
I
n
N
ov
e
m
be
r
2025,
af
t
e
r
a
s
ec
ur
i
ty
b
r
each
i
n
t
he
O
p
e
n
R
e
v
ie
w
syst
e
m
(
w
hich
e
xpos
ed
som
e
r
e
v
ie
w
e
r
ide
nt
i
t
ie
s
f
rom
v
a
r
i
ous
AI
c
on
fe
r
e
n
ce
s
),
t
he
N
e
ur
IPS
c
omm
i
tt
ee
i
ssu
ed
a
n
un
e
qu
i
vo
ca
l
st
a
t
e
m
e
nt
on
p
ee
r
-
r
e
v
ie
w
s
ec
r
ec
y
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
a
nnoun
ce
m
e
nt
r
e
m
i
n
ded
a
ll
p
a
rt
ici
p
a
nts
t
ha
t
l
eaki
n
g
or
e
xplo
i
t
i
n
g
r
e
v
ie
w
i
n
f
orm
a
t
i
on
v
i
ol
a
t
e
s
N
e
ur
IPS
ʼ
s
C
o
de
o
f
C
on
d
u
c
t
a
n
d
c
oul
d
l
ead
to
ba
nn
ed
p
a
rt
ici
p
a
t
i
on
or
r
e
mov
a
l
.
T
he
y
e
st
ab
l
i
s
hed
c
on
fide
nt
ia
l
r
e
port
i
n
g
cha
nn
e
ls
f
or
a
ny
i
n
cide
nts
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
sw
if
t
r
e
spons
e
un
de
rl
i
n
e
s
h
ow
s
e
r
i
ously
N
e
ur
IPS
t
ake
s
r
e
v
ie
w
i
nt
eg
r
i
ty
,
e
sp
ecia
lly
a
s
t
he
c
ommun
i
ty
n
a
v
iga
t
e
s
t
he
op
e
nn
e
ss
o
f
n
e
w
r
e
v
ie
w
i
n
g
pl
a
t
f
orms
.
A
not
he
r
nov
e
l
e
t
hica
l
g
u
ide
l
i
n
e
w
a
s
t
he
2025
LLM P
ol
ic
y
f
or
a
ut
h
ors
a
n
d
r
e
v
ie
w
e
rs
(
n
e
ur
i
ps
.
cc
).
I
t
e
xpl
ici
tly
w
e
l
c
om
e
s
us
e
o
f
AI
/
LLM
s
f
or
wr
i
t
i
n
g
a
ss
i
st
a
n
ce
b
ut
w
i
t
h
c
on
di
t
i
ons
.
A
ut
h
ors
m
a
y
us
e
a
ny
too
l
to
pr
e
p
a
r
e
p
a
p
e
rs
,
b
ut
if
a
n
LLM
i
s
i
nt
eg
r
a
l
to
t
he
m
e
t
h
o
d
(
e
.
g
.
p
a
rt
o
f
t
he
a
l
g
or
i
t
h
m
),
i
ts
us
e
must
be
c
l
ea
rly
d
o
c
um
e
nt
ed
i
n
t
he
p
a
p
e
r
(
n
e
ur
i
ps
.
cc
).
T
ools
us
ed
only
f
or
g
r
a
mm
a
r
or
edi
t
i
n
g
d
o
not
r
e
qu
i
r
e
di
s
c
losur
e
.
C
ru
cia
lly
,
a
ut
h
ors
r
e
m
ai
n
f
ully
r
e
spons
ib
l
e
f
or
a
ll
c
ont
e
nt
:
t
he
y
must
e
nsur
e
c
orr
ec
t
a
n
d
or
igi
n
a
l
t
e
xt
a
n
d
ci
t
a
t
i
ons
(
e
.
g
.
not
ci
t
i
n
g
fake
r
efe
r
e
n
ce
s
ge
n
e
r
a
t
ed
b
y
AI
)
(
n
e
ur
i
ps
.
cc
).
T
he
pol
ic
y
st
a
t
e
s
t
ha
t
m
i
sus
e
(
e
.
g
.
a
ny
fab
r
ica
t
i
ons
b
y
AI
l
ef
t
unv
e
r
ified
)
c
oul
d
l
ead
to
p
a
p
e
r
r
e
tr
ac
t
i
on
e
v
e
n
post
-
c
on
fe
r
e
n
ce
.
F
or
r
e
v
ie
w
e
rs
,
t
he
LLM
pol
ic
y
i
s
a
r
e
m
i
n
de
r
t
ha
t
su
b
m
i
ss
i
ons
must
not
be
di
s
c
uss
ed
w
i
t
h
a
nyon
e
or
a
ny
LLM
,
a
n
d
t
ha
t
su
b
m
i
ss
i
on
c
o
de
a
n
d
c
ont
e
nt
must
be
ke
pt
c
on
fide
nt
ia
l
(
n
e
ur
i
ps
.
cc
).
I
n
s
h
ort
,
N
e
ur
IPS
2025
ba
l
a
n
ced
op
e
nn
e
ss
to
n
e
w
wr
i
t
i
n
g
aid
s
w
i
t
h
str
i
n
ge
nt
v
e
r
ifica
t
i
on
st
a
n
da
r
d
s
.
O
v
e
r
a
ll
,
t
he
2025
r
e
v
ie
w
pro
ce
ss
ca
n
be
cha
r
ac
t
e
r
i
z
ed
a
s
a
s
ca
l
ab
l
e
y
e
t
r
ig
orous
syst
e
m
:
on
e
str
i
v
i
n
g
to
m
ai
nt
ai
n
qu
a
l
i
ty
a
m
id
e
xplos
i
on
.
T
he
chai
rs
ʼ
r
e
trosp
ec
t
i
v
e
s
s
h
ow
c
on
fide
n
ce
t
ha
t
acce
pt
ed
p
a
p
e
rs
m
e
t
N
e
ur
IPS
ba
r
(
out
c
om
e
s
“
on
p
a
r
w
i
t
h
pr
e
v
i
ous
y
ea
rs
”
(
b
lo
g
.
n
e
ur
i
ps
.
cc
))
w
hi
l
e
ca
ll
i
n
g
f
or
on
g
o
i
n
g
i
mprov
e
m
e
nts
(
e
.
g
.
mor
e
r
e
v
ie
w
e
r
di
v
e
rs
i
ty
,
c
l
ea
r
e
r
g
u
ide
l
i
n
e
s
on
da
t
a
d
o
c
um
e
nt
a
t
i
on
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)).
4.
T
ech
n
ica
l
P
ro
g
r
a
m
H
igh
l
igh
ts
N
e
ur
IPS
ʼ
s
t
ech
n
ica
l
c
or
e
i
s
i
ts
p
ee
r
-
r
e
v
ie
w
ed
pro
g
r
a
m
.
H
e
r
e
w
e
e
x
a
m
i
n
e
ke
y
e
l
e
m
e
nts
o
f
N
e
ur
IPS
2025ʼ
s
pro
g
r
a
m
–
b
ot
h
t
he
su
b
st
a
nt
i
v
e
n
e
w
k
nowl
edge
pr
e
s
e
nt
ed
a
n
d
t
he
stru
c
tur
e
o
f
s
e
ss
i
ons
.
4.1
K
e
y
T
he
m
e
s
a
n
d
T
r
e
n
d
s
M
a
ny
e
m
e
r
gi
n
g
t
he
m
e
s
p
e
rv
aded
N
e
ur
IPS
2025.
B
ot
h
t
he
acce
pt
ed
p
a
p
e
rs
a
n
d
i
nv
i
t
ed
t
a
l
k
s
e
mp
ha
s
i
z
ed
la
r
ge
-
s
cale
lea
rn
i
n
g
syst
em
s
,
m
u
l
t
im
o
dal
AI
,
a
n
d
f
oun
da
t
i
on
al
t
he
ory
.
S
om
e
o
f
t
he
d
om
i
n
a
nt
top
ic
s
w
e
r
e
:
L
a
r
ge
L
a
n
g
u
age
a
n
d
F
oun
da
t
i
on
M
o
de
ls
:
G
i
v
e
n
t
he
on
g
o
i
n
g
AI
r
e
volut
i
on
,
m
a
ny
p
a
p
e
rs
f
o
c
us
ed
on
un
de
rst
a
n
di
n
g
or
i
mprov
i
n
g
LLM
s
.
T
he
B
e
st
P
a
p
e
r
(
Q
i
u
e
t
al
.
)
stu
died
ga
t
i
n
g
m
echa
n
i
sms
i
n
tr
a
ns
f
orm
e
r
-
ba
s
ed
LLM
s
to
i
mprov
e
efficie
n
c
y
a
n
d
st
abi
l
i
ty
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
O
t
he
r
p
a
p
e
rs
i
nv
e
st
iga
t
ed
r
ea
son
i
n
g
i
n
LLM
s
(
e
.
g
.
w
he
t
he
r
RL
fi
n
e
-
tun
i
n
g
truly
e
xp
a
n
d
s
t
hei
r
ca
p
abi
l
i
t
ie
s
–
on
e
runn
e
r
-
up
ʼ
s
n
ega
t
i
v
e
fi
n
di
n
g
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)),
i
nt
e
rpr
e
t
abi
l
i
ty
(
e
.
g
.
pro
bi
n
g
r
e
pr
e
s
e
nt
a
t
i
on
ge
om
e
try
(
n
i
ps
.
cc
)),
a
n
d
ro
b
ustn
e
ss
(
e
.
g
.
w
a
t
e
rm
a
r
ki
n
g
v
ec
tor
da
t
aba
s
e
s
(
n
i
ps
.
cc
),
c
ont
e
nt
fi
lt
e
r
i
n
g
).
W
or
k
s
h
ops
on
R
e
spons
ib
l
e
F
oun
da
t
i
on
M
o
de
ls
a
n
d
LLM P
e
rson
a
M
o
de
l
i
n
g
(
s
ee
S
ec
t
i
on
6)
f
urt
he
r
high
l
igh
t
t
he
e
mp
ha
s
i
s
on
t
he
i
mpl
ica
t
i
ons
o
f
t
he
s
e
pow
e
r
f
ul
mo
de
ls
.
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 6 of 16
R
ei
n
f
or
ce
m
e
nt
L
ea
rn
i
n
g
a
n
d
A
ge
nts
:
RL
c
ont
i
nu
e
s
to
be
a
m
aj
or
f
o
c
us
,
b
ot
h
i
n
pur
e
ML
a
n
d
i
n
mult
i
-
age
nt
or
age
nt
ic
c
ont
e
xts
.
F
or
e
x
a
mpl
e
,
on
e
B
e
st
P
a
p
e
r
(“1000
L
a
y
e
r
N
e
twor
k
s
f
or
S
e
l
f
-
S
up
e
rv
i
s
ed
RL
”)
e
x
a
m
i
n
ed
e
xtr
e
m
e
ly
dee
p
n
e
twor
k
s
f
or
RL
g
o
a
l
achie
v
e
m
e
nt
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
runn
e
r
-
up
“
RL R
ea
son
i
n
g
”
p
a
p
e
r
c
r
i
t
ica
lly
a
ss
e
ss
ed
h
ow
RL
fi
n
e
-
tun
i
n
g
w
i
t
h
v
e
r
ifiab
l
e
r
e
w
a
r
d
s
(
RL
V
R
)
e
n
ha
n
ce
s
LLM
p
e
r
f
orm
a
n
ce
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
c
on
c
lu
di
n
g
t
ha
t
c
urr
e
nt
RL
m
e
t
h
o
d
s
m
a
y
not
unlo
ck
f
un
da
m
e
nt
a
lly
n
e
w
abi
l
i
t
ie
s
(
e
n
c
our
agi
n
g
f
utur
e
wor
k
on
mor
e
pow
e
r
f
ul
RL
p
a
r
adig
ms
).
T
utor
ia
ls
a
n
d
wor
k
s
h
ops
,
su
ch
a
s
t
he
C
e
nt
e
r
i
n
g
L
ow
-
R
e
sour
ce
L
a
n
g
u
age
s
wor
k
s
h
op
a
n
d
v
a
r
i
ous
RL
/
age
nts
tutor
ia
ls
,
r
ef
l
ec
t
ed
b
ro
ad
i
nt
e
r
e
st
i
n
de
ploy
i
n
g
AI
age
nts
r
e
spons
ib
ly
.
D
a
t
a
s
e
ts
,
B
e
n
ch
m
a
r
k
s
,
a
n
d
E
v
a
lu
a
t
i
on
:
T
he
da
t
a
s
e
ts
-
ce
ntr
ic
p
e
rsp
ec
t
i
v
e
w
a
s
high
l
igh
t
ed
b
y
t
he
dedica
t
ed
tr
ack
.
A
cce
pt
ed
wor
k
s
o
f
t
e
n
i
ntro
d
u
ced
n
e
w
be
n
ch
m
a
r
k
s
f
or
sp
ecia
l
i
z
ed
d
om
ai
ns
(
e
.
g
.
a
be
n
ch
m
a
r
k
f
or
o
cea
n
f
or
eca
st
i
n
g
(
n
i
ps
.
cc
),
c
l
i
m
a
t
e
or
E
a
rt
h
da
t
a
t
a
s
k
s
)
or
m
e
tr
ic
s
f
or
mo
de
l
e
v
a
lu
a
t
i
on
.
T
he
D
&
B
b
lo
g
r
e
port
ed
t
ha
t
84
%
o
f
acce
pt
ed
D
&
B
p
a
p
e
rs
pr
e
s
e
nt
ed
n
e
w
da
t
a
s
e
ts
to
e
xp
a
n
d
e
v
a
lu
a
t
i
on
poss
ibi
l
i
t
ie
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
m
i
rrors
c
ommun
i
ty
c
on
ce
rns
ab
out
e
v
a
lu
a
t
i
n
g
mo
de
ls
fai
rly
on
n
e
w
t
a
s
k
s
.
C
ro
i
ss
a
nt
m
e
t
ada
t
a
eff
orts
s
h
ow
a
n
i
n
c
r
ea
s
i
n
g
pus
h
f
or
tr
a
nsp
a
r
e
n
c
y
i
n
da
t
a
c
ur
a
t
i
on
.
A
ppl
ica
t
i
ons
a
n
d
I
nt
e
r
di
s
ci
pl
i
n
a
ry
AI
:
M
a
ny
p
a
p
e
rs
t
ack
l
ed
AI
f
or
s
cie
n
ce
,
hea
lt
hca
r
e
,
or
so
cia
l
g
oo
d
.
E
x
a
mpl
e
top
ic
s
i
n
c
lu
ded
p
h
ys
ic
s
-
ba
s
ed
mo
de
l
i
n
g
,
ro
b
ot
ic
p
e
r
ce
pt
i
on
(
e
.
g
.
“
RAPID H
a
n
d
”
f
or
low
-
c
ost
de
xt
e
rous
m
a
n
i
pul
a
t
i
on
(
n
i
ps
.
cc
)),
bi
olo
g
y
a
n
d
m
edici
n
e
(
pr
edic
t
i
n
g
prot
ei
n
s
e
qu
e
n
ce
s
,
d
ru
g
di
s
c
ov
e
ry
),
a
n
d
ec
onom
e
tr
ic
s
.
T
h
r
ead
s
on
so
cia
l
r
e
spons
ibi
l
i
ty
r
a
n
t
h
rou
gh
out
:
wor
k
s
h
ops
on
so
cia
l
trust
a
n
d
pol
ic
ym
aki
n
g
,
pos
i
t
i
on
p
a
p
e
rs
,
a
n
d
i
nv
i
t
ed
t
a
l
k
s
(
e
.
g
.
Z
e
yn
e
p
T
u
fekci
on
so
cie
t
a
l
i
mp
ac
ts
o
f
“
g
oo
d
-
e
nou
gh
”
AI
(
b
lo
g
.
n
e
ur
i
ps
.
cc
))
un
de
rs
c
or
ed
e
t
hica
l
a
n
d
r
ea
l
-
worl
d
di
m
e
ns
i
ons
.
T
he
s
e
tr
e
n
d
s
a
r
e
not
surpr
i
s
i
n
g
,
gi
v
e
n
t
he
t
i
m
i
n
g
.
T
he
i
nv
i
t
ed
sp
eake
r
l
i
n
e
up
r
ef
l
ec
t
ed
t
he
s
e
t
he
m
e
s
:
R
icha
r
d
S
utton
(
D
ec
3,
morn
i
n
g
)
–
fa
m
ed
RL
p
i
on
ee
r
spo
ke
c
r
i
t
ica
lly
ab
out
t
he
tr
ajec
tory
o
f
AI
:
“
a
s
AI
ha
s
bec
om
e
a
h
u
ge
i
n
d
ustry
,
to
a
n
e
xt
e
nt
i
t
ha
s
lost
i
ts
w
a
y
,”
he
a
r
g
u
ed
,
ad
vo
ca
t
i
n
g
a
r
e
turn
to
l
ea
rn
i
n
g
age
nts
t
ha
t
b
u
i
l
d
worl
d
mo
de
ls
a
n
d
l
ea
rn
c
ont
i
nu
a
lly
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
ca
ll
to
f
un
da
m
e
nt
a
ls
ech
o
e
s
t
he
or
i
sts
ʼ
c
on
ce
rns
a
m
id
t
he
r
ece
nt
f
o
c
us
on
s
ca
l
e
.
Z
e
yn
e
p
T
u
fekci
(
D
ec
3,
af
t
e
rnoon
)
–
a
so
ci
olo
gi
st
o
f
t
ech
nolo
g
y
,
s
he
w
a
rn
ed
t
ha
t
“
A
rt
ificia
l
G
oo
d
-
E
nou
gh
I
nt
e
ll
ige
n
ce
”
i
s
a
lr
ead
y
di
srupt
i
v
e
e
v
e
n
ab
s
e
nt
AGI
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
H
e
r
t
a
l
k
e
mp
ha
s
i
z
ed
t
ha
t
ge
n
e
r
a
t
i
v
e
AI
ʼ
s
r
ea
l
i
mp
ac
t
c
om
e
s
f
rom
m
aki
n
g
e
x
i
st
i
n
g
t
a
s
k
s
chea
p
a
n
d
s
ca
l
ab
l
e
,
w
hich
r
i
s
k
s
un
de
rm
i
n
i
n
g
our
syst
e
ms
o
f
trut
h
a
n
d
a
ut
he
nt
ici
ty
.
Y
eji
n
C
h
o
i
(
D
ec
4,
morn
i
n
g
,
P
osn
e
r
L
ec
tur
e
)
–
s
he
add
r
e
ss
ed
“
jagged
i
nt
e
ll
ige
n
ce
”
i
n
NLP
mo
de
ls
:
de
sp
i
t
e
be
n
ch
m
a
r
k
pro
g
r
e
ss
,
mo
de
ls
st
i
ll
e
x
hibi
t
i
n
c
ons
i
st
e
nt
r
ea
son
i
n
g
a
n
d
c
ontr
adic
t
i
ons
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
C
h
o
i
str
e
ss
ed
ga
ps
i
n
c
ommons
e
ns
e
a
n
d
t
he
n
eed
f
or
be
tt
e
r
AI
un
de
rp
i
nn
i
n
g
s
.
M
e
l
a
n
ie
M
i
t
che
ll
(
D
ec
4,
af
t
e
rnoon
)
–
a
c
o
g
n
i
t
i
v
e
s
cie
nt
i
st
,
s
he
qu
e
st
i
on
ed
LLM
c
o
g
n
i
t
i
on
:
mo
de
ls
surp
a
ss
h
um
a
ns
on
m
a
ny
t
e
sts
b
ut
fai
l
“
un
h
um
a
nl
ike
w
a
ys
.”
U
s
i
n
g
m
e
t
h
o
d
s
f
rom
c
omp
a
r
a
t
i
v
e
psy
ch
olo
g
y
(
a
n
a
lo
gie
s
,
v
i
su
a
l
ab
str
ac
t
i
on
t
a
s
k
s
),
s
he
pr
e
s
e
nt
ed
n
e
w
a
ppro
ache
s
to
e
v
a
lu
a
t
e
AI
c
o
g
n
i
t
i
on
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
K
yun
gh
yun
C
h
o
(
D
ec
5,
morn
i
n
g
)
–
a
dee
p
l
ea
rn
i
n
g
r
e
s
ea
r
che
r
,
C
h
o
surv
e
y
ed
hi
s
ca
r
ee
r
-
sp
a
nn
i
n
g
pro
b
l
e
ms
(
GAN
s
,
tr
a
nsl
a
t
i
on
,
prot
ei
ns
),
r
ef
l
ec
t
i
n
g
on
pro
b
l
e
m
f
ormul
a
t
i
on
i
n
AI
a
n
d
h
ow
di
v
e
rs
e
t
a
s
k
s
a
r
e
i
nt
e
r
c
onn
ec
t
ed
f
rom
t
ech
n
ica
l
,
so
cia
l
,
a
n
d
p
e
rson
a
l
p
e
rsp
ec
t
i
v
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
n
d
r
e
w
S
a
x
e
(
D
ec
5,
af
t
e
rnoon
,
B
r
ei
m
a
n
L
ec
tur
e
)
–
he
pr
e
s
e
nt
ed
m
a
t
he
m
a
t
ica
l
a
n
a
lys
e
s
o
f
h
ow
n
e
twor
k
a
r
chi
t
ec
tur
e
,
da
t
a
,
a
n
d
i
n
i
t
ia
l
i
z
a
t
i
on
i
nt
e
r
ac
t
to
y
ie
l
d
c
ompl
e
x
r
e
pr
e
s
e
nt
a
t
i
ons
.
S
a
x
e
i
ntro
d
u
ced
a
“
n
e
ur
a
l
r
ace
”
m
e
t
a
p
h
or
w
he
r
e
diffe
r
e
nt
p
a
t
h
w
a
ys
i
n
a
n
e
twor
k
c
omp
e
t
e
to
e
xpl
ai
n
da
t
a
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
H
i
s
t
a
l
k
s
hed
l
igh
t
on
hidde
n
stru
c
tur
e
s
g
ov
e
rn
i
n
g
l
ea
rn
i
n
g
a
n
d
ge
n
e
r
a
l
i
z
a
t
i
on
.
I
n
summ
a
ry
,
i
nv
i
t
ed
t
a
l
k
s
sp
a
nn
ed
f
oun
da
t
i
on
a
l
deba
t
e
s
(
S
utton
,
C
h
o
i
),
so
cie
t
a
l
c
r
i
t
i
qu
e
(
T
u
fekci
),
c
o
g
n
i
t
i
v
e
e
v
a
lu
a
t
i
on
o
f
AI
(
M
i
t
che
ll
),
i
nt
e
r
di
s
ci
pl
i
n
a
ry
p
e
rsp
ec
t
i
v
e
(
C
h
o
),
a
n
d
t
he
or
e
t
ica
l
a
n
a
lys
i
s
(
S
a
x
e
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
ec
l
ec
t
ic
s
e
t
to
ge
t
he
r
high
l
igh
t
ed
w
he
r
e
N
e
ur
IPS
2025
s
i
tu
a
t
e
s
:
a
t
t
he
i
nt
e
rs
ec
t
i
on
o
f
c
utt
i
n
g
-
edge
e
n
gi
n
ee
r
i
n
g
a
n
d
i
ntrosp
ec
t
i
v
e
s
cie
n
ce
.
4.2
N
ot
ab
l
e
P
a
p
e
r
E
x
a
mpl
e
s
T
he
t
ech
n
ica
l
c
ontr
ib
ut
i
ons
a
t
N
e
ur
IPS
2025
c
ov
e
r
ed
a
v
a
st
a
rr
a
y
o
f
top
ic
s
.
A
mon
g
t
h
ous
a
n
d
s
o
f
p
a
p
e
rs
,
som
e
e
x
e
mpl
if
y
t
he
c
utt
i
n
g
edge
:
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 7 of 16
G
a
t
ed
A
tt
e
nt
i
on
i
n
LLM
s
(
B
e
st
P
a
p
e
r
)
–
Q
i
u
e
t
al
.
i
ntro
d
u
ced
a
s
i
mpl
e
y
e
t
pow
e
r
f
ul
mo
difica
t
i
on
to
T
r
a
ns
f
orm
e
r
a
tt
e
nt
i
on
:
i
ns
e
rt
i
n
g
a
head
-
sp
ecific
s
igm
o
id
ga
t
e
af
t
e
r
t
he
st
a
n
da
r
d
so
f
tm
a
x
op
e
r
a
t
i
on
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
c
ross
e
xp
e
r
i
m
e
nts
on
1.7
B
a
n
d
15
B
mo
de
ls
(
tr
ai
n
ed
on
tr
i
ll
i
ons
o
f
to
ke
ns
),
t
hi
s
ga
t
i
n
g
i
mprov
ed
st
abi
l
i
ty
a
n
d
lon
g
-
c
ont
e
xt
p
e
r
f
orm
a
n
ce
,
a
n
d
m
i
t
iga
t
ed
“
a
tt
e
nt
i
on
s
i
n
k
”
i
ssu
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
B
e
st
P
a
p
e
r
C
omm
i
tt
ee
l
a
u
ded
i
t
f
or
de
l
i
v
e
r
i
n
g
a
str
aigh
t
f
orw
a
r
d
cha
n
ge
w
i
t
h
b
ro
ad
i
mp
ac
t
,
l
ike
ly
to
be
w
ide
ly
ad
opt
ed
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
(
N
ot
ab
ly
,
t
he
w
i
nn
i
n
g
wor
k
ca
m
e
f
rom
a
n
i
n
d
ustry
l
ab
,
r
ef
l
ec
t
i
n
g
t
he
e
normous
r
e
sour
ce
s
now
d
r
i
v
i
n
g
ML
r
e
s
ea
r
ch
.)
D
ee
p
RL S
ca
l
i
n
g
(
R
unn
e
r
-
U
p
)
–
W
a
n
g
e
t
al
.
e
xplor
ed
e
xtr
e
m
e
ly
dee
p
n
e
twor
k
s
f
or
s
e
l
f
-
sup
e
rv
i
s
ed
RL
,
s
h
ow
i
n
g
t
ha
t
“1000-
l
a
y
e
r
”
a
r
chi
t
ec
tur
e
s
ca
n
a
tt
ai
n
n
e
w
ca
p
abi
l
i
t
ie
s
i
n
g
o
a
l
-
r
eachi
n
g
t
a
s
k
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
lt
h
ou
gh
v
e
ry
dee
p
,
t
he
s
e
n
e
twor
k
s
r
e
t
ai
n
ed
l
ea
rn
abi
l
i
ty
a
n
d
a
llow
ed
s
h
ow
i
n
g
h
ow
de
pt
h
i
n
f
lu
e
n
ce
s
RL
p
e
r
f
orm
a
n
ce
.
RL
f
or
LLM R
ea
son
i
n
g
(
R
unn
e
r
-
U
p
)
–
Y
u
e
e
t
al
.
p
e
r
f
orm
ed
a
r
ig
orous
i
nv
e
st
iga
t
i
on
i
nto
R
ei
n
f
or
ce
m
e
nt
L
ea
rn
i
n
g
w
i
t
h
V
e
r
ifiab
l
e
R
e
w
a
r
d
s
(
RL
V
R
).
C
ontr
a
ry
to
pr
e
v
ai
l
i
n
g
a
ssumpt
i
ons
,
t
he
y
f
oun
d
t
ha
t
RL
V
R
d
o
e
s
not
pro
d
u
ce
f
un
da
m
e
nt
a
lly
n
e
w
r
ea
son
i
n
g
i
n
LLM
s
be
yon
d
t
hei
r
ba
s
e
mo
de
ls
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hei
r
a
n
a
lys
i
s
ac
ross
mult
i
pl
e
mo
de
l
fa
m
i
l
ie
s
a
n
d
t
a
s
k
s
s
h
ow
ed
t
ha
t
RL
m
ai
nly
a
mpl
ifie
s
e
x
i
st
i
n
g
c
orr
ec
t
beha
v
i
ors
r
a
t
he
r
t
ha
n
e
xp
a
n
di
n
g
t
he
solut
i
on
sp
ace
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
n
ega
t
i
v
e
r
e
sult
(
a
n
d
i
ts
hi
stor
ica
l
p
e
rsp
ec
t
i
v
e
)
w
a
s
dee
m
ed
“
c
r
i
t
ica
lly
i
mport
a
nt
”
b
y
r
e
v
ie
w
e
rs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
l
ike
ly
prompt
i
n
g
f
utur
e
r
e
s
ea
r
ch
i
nto
mor
e
pow
e
r
f
ul
RL
p
a
r
adig
ms
.
T
he
or
e
t
ica
l
A
d
v
a
n
ce
s
–
A
su
i
t
e
o
f
n
e
w
t
he
or
e
t
ica
l
i
ns
igh
ts
w
e
r
e
a
lso
ce
l
eb
r
a
t
ed
.
F
or
e
x
a
mpl
e
,
C
ha
s
e
e
t
al
.
prov
ed
t
igh
t
m
i
st
ake
b
oun
d
s
i
n
tr
a
ns
d
u
c
t
i
v
e
l
ea
rn
i
n
g
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
a
“30-
y
ea
r
-
ol
d
op
e
n
pro
b
l
e
m
”),
ea
rn
i
n
g
runn
e
r
-
up
di
st
i
n
c
t
i
on
f
or
i
ts
b
r
eak
t
h
rou
gh
on
unl
abe
l
ed
da
t
a
be
n
efi
ts
.
M
ea
nw
hi
l
e
,
a
h
ost
o
f
acce
pt
ed
p
a
p
e
rs
de
lv
ed
i
nto
opt
i
m
i
z
a
t
i
on
t
he
ory
,
ge
n
e
r
a
l
i
z
a
t
i
on
b
oun
d
s
,
a
n
d
n
e
w
a
l
g
or
i
t
h
ms
(
only
tourn
a
m
e
nts
o
f
su
ch
t
i
tl
e
s
a
r
e
v
i
s
ib
l
e
i
n
t
he
o
fficia
l
pro
g
r
a
m
l
i
st
(
n
i
ps
.
cc
)).
A
ppl
ied
a
n
d
D
om
ai
n
-
S
p
ecific
W
or
k
–
M
a
ny
p
a
p
e
rs
add
r
e
ss
ed
pr
ac
t
ica
l
d
om
ai
ns
.
F
or
i
nst
a
n
ce
,
H
a
n
e
t
al
.
m
a
y
ha
v
e
pr
e
s
e
nt
ed
“
O
cea
n
B
e
n
ch
”
(
be
n
ch
m
a
r
ki
n
g
g
lo
ba
l
o
cea
n
f
or
eca
st
i
n
g
AI
)
(
n
i
ps
.
cc
);
C
he
e
t
al
.
c
ontr
ib
ut
ed
pr
i
v
ac
y
-
pr
e
s
e
rv
i
n
g
fede
r
a
t
ed
l
ea
rn
i
n
g
m
e
t
h
o
d
s
;
ro
b
ot
ic
s
r
e
s
ea
r
che
rs
i
ntro
d
u
ced
n
e
w
da
t
a
s
e
ts
f
or
e
m
b
o
died
syst
e
ms
.
W
or
k
s
h
ops
i
n
M
e
x
ic
o
C
i
ty
t
a
r
ge
t
ed
r
ea
l
-
worl
d
c
on
ce
rns
(
e
.
g
.
de
ploy
i
n
g
v
i
s
i
on
-
l
a
n
g
u
age
mo
de
ls
ro
b
ustly
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)).
T
he
b
r
ead
t
h
w
a
s
e
normous
,
f
rom
a
strop
h
ys
ic
s
to
ag
r
ic
ultur
e
,
un
de
rs
c
or
i
n
g
N
e
ur
IPS
ʼ
s
rol
e
a
s
a
pl
a
t
f
orm
f
or
b
ot
h
c
or
e
ML
a
n
d
i
ts
a
ppl
ica
t
i
ons
.
C
oll
ec
t
i
v
e
ly
,
t
he
t
ech
n
ica
l
c
ontr
ib
ut
i
ons
o
f
N
e
ur
IPS
2025
r
ef
l
ec
t
a
m
a
tur
i
n
g
fie
l
d
g
r
a
ppl
i
n
g
w
i
t
h
s
ca
l
e
,
r
e
pro
d
u
cibi
l
i
ty
,
a
n
d
i
mp
ac
t
.
LLM
s
a
n
d
RL
a
r
e
m
aj
or
wor
ki
n
g
h
ors
e
s
,
b
ut
t
he
r
e
i
s
a
lso
c
ont
i
nu
ed
a
tt
e
nt
i
on
to
d
om
ai
ns
l
ike
i
nt
e
rpr
e
t
abi
l
i
ty
,
fai
rn
e
ss
,
a
n
d
r
ea
l
-
worl
d
syst
e
ms
.
D
e
t
ai
l
ed
st
a
t
i
st
ic
s
o
f
top
ic
s
w
i
ll
e
m
e
r
ge
i
n
c
on
fe
r
e
n
ce
pro
ceedi
n
g
s
(
e
.
g
.
ke
ywor
d
m
e
t
a
-
a
n
a
lys
i
s
),
b
ut
e
v
e
n
t
hi
s
sn
a
ps
h
ot
s
h
ows
t
he
c
on
fe
r
e
n
ce
m
i
rrors
t
he
fie
l
d
ʼ
s
pr
i
or
i
t
ie
s
:
ge
n
e
r
a
t
i
v
e
AI
,
da
t
a
/
be
n
ch
m
a
r
k
s
,
r
ig
orous
e
v
a
lu
a
t
i
on
,
a
n
d
c
ross
-
di
s
ci
pl
i
n
a
ry
syn
e
r
g
y
.
4.3
J
ourn
a
ls
a
n
d
C
ross
-
D
i
s
ci
pl
i
n
e
B
r
idgi
n
g
I
n
a
n
i
nnov
a
t
i
v
e
e
xp
e
r
i
m
e
nt
,
N
e
ur
IPS
2025
b
r
idged
to
j
ourn
al
pu
bli
s
hi
n
g
.
A
s
de
s
c
r
ibed
i
n
t
he
N
e
ur
IPS
b
lo
g
,
t
he
c
on
fe
r
e
n
ce
i
nv
i
t
ed
top
r
ece
nt
j
ourn
a
l
p
a
p
e
rs
f
rom
ML
a
n
d
st
a
ts
to
pr
e
s
e
nt
a
s
post
e
rs
.
T
he
i
n
a
u
g
ur
a
l
J
ourn
a
l
T
r
ack
fea
tur
ed
34
p
a
p
e
rs
(14
f
rom
J
ourn
al
o
f
M
achi
n
e
L
ea
rn
i
n
g
R
e
s
ea
r
ch
a
n
d
20
f
rom
A
nn
al
s
o
f
S
t
a
t
i
st
ic
s
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
s
e
s
e
l
ec
t
i
ons
i
n
c
lu
ded
ad
v
a
n
ced
t
he
or
e
t
ica
l
a
n
d
m
e
t
h
o
d
olo
gica
l
wor
k
c
ons
i
st
e
nt
w
i
t
h
N
e
ur
IPS
v
a
lu
e
s
.
F
or
e
x
a
mpl
e
,
f
rom
JMLR
ca
m
e
p
a
p
e
rs
on
ro
b
ust
RL
opt
i
m
a
a
n
d
sp
ec
tr
a
l
m
e
t
h
o
d
s
;
f
rom
A
o
S
ca
m
e
p
a
p
e
rs
on
pr
i
v
ac
y
a
n
d
ge
n
e
r
a
l
i
z
a
t
i
on
.
T
he
g
o
a
l
w
a
s
to
high
l
igh
t
“
t
he
highe
st
st
a
n
da
r
d
s
o
f
s
ch
ol
a
rs
hi
p
ac
ross
t
he
ory
,
m
e
t
h
o
d
olo
g
y
,
a
n
d
a
ppl
ica
t
i
ons
”
a
n
d
to
f
ost
e
r
i
nt
e
r
ac
t
i
on
be
tw
ee
n
N
e
ur
IPS
a
tt
e
n
dee
s
a
n
d
t
he
b
ro
ade
r
acade
m
ic
l
i
t
e
r
a
tur
e
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
I
n
pr
ac
t
ice
,
t
he
J
ourn
a
l
T
r
ack
a
u
g
m
e
nts
t
he
pro
g
r
a
m
w
i
t
h
di
v
e
rs
e
p
e
rsp
ec
t
i
v
e
s
typ
ica
lly
ab
s
e
nt
f
rom
c
on
fe
r
e
n
ce
dead
l
i
n
e
s
.
I
t
e
x
e
mpl
ifie
s
N
e
ur
IPS
ʼ
s
i
n
c
r
ea
s
i
n
g
l
i
n
kage
w
i
t
h
t
he
st
a
t
i
st
ic
s
c
ommun
i
ty
a
n
d
r
ec
o
g
n
i
t
i
on
o
f
r
ig
orous
,
p
ee
r
-
r
e
v
ie
w
ed
c
ontr
ib
ut
i
ons
t
ha
t
m
a
y
not
e
m
e
r
ge
v
ia
c
onv
e
nt
i
on
a
l
ML
c
on
fe
r
e
n
ce
rout
e
s
.
5.
C
ommun
i
ty
a
n
d
S
o
cia
l
A
sp
ec
ts
N
e
ur
IPS
,
be
yon
d
t
ech
n
ica
l
p
a
p
e
rs
,
i
s
a
so
cia
l
e
v
e
nt
f
or
t
he
AI
c
ommun
i
ty
.
S
e
v
e
r
a
l
i
n
i
t
ia
t
i
v
e
s
a
t
N
e
ur
IPS
2025
f
ost
e
r
ed
c
oll
ab
or
a
t
i
on
,
i
n
c
lus
i
on
,
a
n
d
n
e
twor
ki
n
g
.
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 8 of 16
5.1
A
ffi
n
i
ty
a
n
d
S
o
cia
l
E
v
e
nts
T
o
promot
e
un
de
rr
e
pr
e
s
e
nt
ed
g
roups
a
n
d
c
ommun
i
ty
b
u
i
l
di
n
g
,
N
e
ur
IPS
c
ont
i
nu
ed
i
ts
affi
n
i
ty
e
v
e
nts
pro
g
r
a
m
.
I
n
2025
t
he
r
e
w
e
r
e
7
o
fficia
l
affi
n
i
ty
e
v
e
nts
(
wor
k
s
h
ops
or
post
e
r
s
e
ss
i
ons
)
c
o
-
lo
ca
t
ed
w
i
t
h
t
he
c
on
fe
r
e
n
ce
(
n
e
ur
i
ps
.
cc
)
(
n
e
ur
i
ps
.
cc
).
T
he
s
e
i
n
c
lu
ded
:
W
o
me
n
i
n
M
achi
n
e
L
ea
rn
i
n
g
,
L
a
t
i
n
X
i
n
AI
,
Q
u
ee
r
i
n
AI
,
I
n
dige
nous
i
n
AI
/
ML
,
M
us
lim
s
i
n
ML
,
N
e
w
i
n
ML
(
f
or
n
e
w
c
om
e
rs
),
a
n
d
a
j
o
i
nt
post
e
r
s
e
ss
i
on
.
(
T
he
Q
u
ee
r
i
n
AI
e
v
e
nt
,
f
or
e
x
a
mpl
e
,
r
a
n
D
ec
4
a
n
d
w
e
l
c
om
ed
i
n
c
lus
i
v
e
p
a
rt
ici
p
a
t
i
on
.)
T
he
affi
l
ia
t
i
ons
r
ef
l
ec
t
a
b
ro
ad
sp
ec
trum
o
f
c
ommun
i
t
ie
s
.
E
ach
e
v
e
nt
fea
tur
ed
t
a
l
k
s
,
p
a
n
e
ls
,
a
n
d
so
cia
l
i
nt
e
r
ac
t
i
ons
;
f
or
i
nst
a
n
ce
,
t
he
M
usl
i
ms
i
n
ML
wor
k
s
h
op
i
n
c
lu
ded
i
nv
i
t
ed
t
a
l
k
s
a
n
d
m
e
ntor
i
n
g
,
ai
m
i
n
g
to
c
onn
ec
t
M
usl
i
m
AI
r
e
s
ea
r
che
rs
g
lo
ba
lly
(
n
e
ur
i
ps
.
cc
).
S
i
m
i
l
a
rly
,
“
N
e
ur
IPS S
o
cia
ls
”
w
e
r
e
i
n
f
orm
a
l
ga
t
he
r
i
n
g
s
a
n
d
di
s
c
uss
i
ons
l
ed
b
y
c
ommun
i
ty
m
e
m
be
rs
on
i
nt
e
r
di
s
ci
pl
i
n
a
ry
top
ic
s
.
N
e
ur
IPS
2025
s
ched
ul
ed
8
acce
pt
ed
so
cia
l
e
v
e
nts
(
i
n
addi
t
i
on
to
t
he
s
cie
nt
ific
pro
g
r
a
m
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
s
e
r
a
n
ged
f
rom
t
ech
n
ica
l
to
c
r
ea
t
i
v
e
:
high
l
igh
ts
i
n
c
lu
ded
(
i
)
“
L
ea
rn
i
n
g
T
he
ory
A
llia
n
ce
”
–
a
fi
r
e
s
ide
cha
t
a
n
d
n
e
twor
ki
n
g
f
or
t
he
or
i
sts
,
w
hich
ai
m
ed
to
str
e
n
g
t
he
n
t
ie
s
be
tw
ee
n
t
he
ory
a
n
d
a
ppl
ied
ML
(
b
lo
g
.
n
e
ur
i
ps
.
cc
);
(
ii
)
“
W
he
n
E
rrors
D
r
eam
”
–
a
c
r
ea
t
i
v
e
wor
k
s
h
op
r
ef
r
a
m
i
n
g
AI
ha
llu
ci
n
a
t
i
ons
a
s
a
rt
,
w
he
r
e
p
a
rt
ici
p
a
nts
c
o
-
c
r
ea
t
ed
i
nt
e
r
ac
t
i
v
e
mult
i
m
edia
us
i
n
g
AI
outputs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
);
(
iii
)
“
AI
i
n
S
cie
nt
ific
P
ee
r
R
e
v
ie
w
”
–
a
p
a
n
e
l
di
s
c
uss
i
on
e
xplor
i
n
g
AI
ʼ
s
rol
e
i
n
pu
b
l
i
s
hi
n
g
e
t
hic
s
a
n
d
tr
a
nsp
a
r
e
n
c
y
(
b
lo
g
.
n
e
ur
i
ps
.
cc
);
(
i
v
)
“
N
e
uro
AI
”
–
a
f
orum
l
i
n
ki
n
g
n
e
uros
cie
n
ce
a
n
d
ML
f
or
ed
u
ca
t
i
on
a
n
d
r
e
s
ea
r
ch
(
b
lo
g
.
n
e
ur
i
ps
.
cc
);
a
n
d
(
v
)
low
e
r
-
pro
fi
l
e
m
ee
tups
on
top
ic
s
l
ike
age
nt
s
afe
ty
a
n
d
op
e
n
da
t
a
.
T
o
ge
t
he
r
,
t
he
so
cia
l
e
v
e
nts
c
ov
e
r
ed
di
v
e
rs
i
ty
o
f
i
nt
e
r
e
sts
(
t
he
ory
,
a
rt
,
e
t
hic
s
,
a
ppl
ica
t
i
on
)
a
n
d
prov
ided
v
e
nu
e
s
f
or
i
n
f
orm
a
l
dia
lo
g
u
e
.
(
S
ee
T
ab
l
e
2
f
or
a
s
e
l
ec
t
i
on
o
f
t
he
s
e
.)
E
v
e
nt
(
S
o
cia
l
/
A
ffi
n
i
ty
)
T
he
m
e
/
P
urpos
e
W
om
e
n
i
n
M
achi
n
e
L
ea
rn
i
n
g
(
A
ffi
n
i
ty
)
M
e
ntor
i
n
g
a
n
d
n
e
twor
ki
n
g
f
or
wom
e
n
i
n
ML
;
p
a
n
e
ls
on
ca
r
ee
r
de
v
e
lopm
e
nt
.
L
a
t
i
n
X
i
n
AI
(
A
ffi
n
i
ty
)
S
h
ow
ca
s
i
n
g
L
a
t
i
n
A
m
e
r
ica
n
AI
r
e
s
ea
r
ch
,
f
ost
e
r
i
n
g
r
egi
on
a
l
c
oll
ab
or
a
t
i
on
.
L
ea
rn
i
n
g
T
he
ory
A
ll
ia
n
ce
(
S
o
cial
)
F
i
r
e
s
ide
cha
t
+
di
s
c
uss
i
ons
on
t
he
ory
i
n
ML
;
m
e
ntors
hi
p
f
or
t
he
or
i
sts
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
W
he
n
E
rrors
D
r
ea
m
(
S
o
cial
)
C
r
ea
t
i
v
e
ja
m
e
xplor
i
n
g
AI
“
ha
llu
ci
n
a
t
i
ons
”
a
s
c
oll
ab
or
a
t
i
v
e
a
rt
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
AI
i
n
S
cie
nt
ific
P
ee
r
R
e
v
ie
w
(
S
o
cial
)
P
a
n
e
l
on
h
ow
AI
ca
n
i
mprov
e
s
ch
ol
a
rly
pu
b
l
i
s
hi
n
g
(
a
ut
h
or
i
n
g
to
r
e
v
ie
w
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
N
e
uro
AI
:
F
rom
N
e
urons
to
T
r
a
ns
f
orm
e
rs
(
S
o
cial
)
I
nt
e
r
di
s
ci
pl
i
n
a
ry
f
orum
on
n
e
uros
cie
n
ce
+
AI
,
ed
u
ca
t
i
on
&
e
t
hic
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
M
usl
i
ms
i
n
ML
(
A
ffi
n
i
ty
)
W
or
k
s
h
op
f
or
M
usl
i
m
ML
r
e
s
ea
r
che
rs
;
t
a
l
k
s
+
m
e
ntor
i
n
g
(
n
e
ur
i
ps
.
cc
).
A
ge
nts
S
afe
ty
P
a
n
e
l
(
S
o
cial
)
P
a
n
e
l
a
n
d
di
s
c
uss
i
on
on
s
afe
ty
o
f
AI
age
nts
(
h
ost
ed
b
y
CAIS
/
AISI
).
T
able
2
:
S
elec
t
ed
affi
n
i
ty
a
n
d
so
cial
e
v
e
nts
a
t
N
e
ur
IP
S
2025
.
T
he
s
e
e
v
e
nts
,
or
ga
n
i
z
ed
b
y
c
ommun
i
ty
m
e
m
be
rs
,
f
ost
e
r
ed
i
n
c
lus
i
on
a
n
d
dia
lo
g
u
e
.
(
D
e
s
c
r
i
pt
i
ons
synt
he
s
i
z
ed
f
rom
o
fficia
l
a
nnoun
ce
m
e
nts
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
n
e
ur
i
ps
.
cc
).)
T
he
stron
g
pr
e
s
e
n
ce
o
f
affi
n
i
ty
e
v
e
nts
i
n
dica
t
e
s
N
e
ur
IPS
ʼ
s
c
ont
i
nu
i
n
g
e
mp
ha
s
i
s
on
di
v
e
rs
i
ty
a
n
d
c
ommun
i
ty
support
.
M
or
e
ov
e
r
,
N
e
ur
IPS F
oun
da
t
i
on
ʼ
s
“
B
r
idgi
n
g
t
he
F
utur
e
”
g
r
a
nts
pro
g
r
a
m
(
a
nnoun
ced
O
c
t
2025
(
b
lo
g
.
n
e
ur
i
ps
.
cc
))
–
f
un
di
n
g
up
to
$
50
K
f
or
i
nnov
a
t
i
v
e
ed
u
ca
t
i
on
/
outr
each
pro
jec
ts
–
f
urt
he
r
s
h
ows
c
omm
i
tm
e
nt
to
b
ro
ade
n
i
n
g
p
a
rt
ici
p
a
t
i
on
i
n
AI
.
5.2
W
or
k
s
h
ops
a
n
d
T
utor
ia
ls
S
a
n
D
ieg
o
W
or
k
s
h
ops
.
T
he
m
ai
n
c
on
fe
r
e
n
ce
i
n
S
a
n
D
ieg
o
h
ost
ed
d
oz
e
ns
o
f
wor
k
s
h
ops
(
typ
ica
lly
1-2
da
y
dee
p
-
di
v
e
s
)
on
sp
ecia
l
i
z
ed
r
e
s
ea
r
ch
a
r
ea
s
.
A
lt
h
ou
gh
a
c
ompl
e
t
e
l
i
st
i
s
l
a
r
ge
,
top
ic
s
i
n
c
lu
ded
:
ge
n
e
r
a
t
i
v
e
mo
de
ls
,
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 9 of 16
ro
b
ot
ic
s
,
s
ec
ur
i
ty
,
fai
rn
e
ss
,
ML
f
or
s
cie
n
ce
,
r
ei
n
f
or
ce
m
e
nt
l
ea
rn
i
n
g
,
n
e
uro
-
AI
,
a
n
d
mor
e
.
M
a
ny
w
e
r
e
i
n
d
ustry
-
sponsor
ed
or
c
o
-
h
ost
ed
(
e
.
g
.
V
i
s
i
on
L
a
n
g
u
age
M
o
del
s
:
R
eal
-
W
or
ld
D
e
p
l
oy
me
nt
,
run
b
y
i
n
d
ustry
pr
ac
t
i
t
i
on
e
rs
).
T
utor
ia
ls
l
ike
w
i
s
e
c
ov
e
r
ed
b
ot
h
f
oun
da
t
i
on
a
l
ML
(
e
.
g
.
B
a
y
e
s
ia
n
dee
p
l
ea
rn
i
n
g
,
P
ro
babi
l
i
st
ic
I
n
fe
r
e
n
ce
)
a
n
d
tr
e
n
di
n
g
su
bjec
ts
(
e
.
g
.
efficie
nt
tr
a
ns
f
orm
e
rs
,
ge
om
e
tr
ic
dee
p
l
ea
rn
i
n
g
).
M
e
x
ic
o
C
i
ty
W
or
k
s
h
ops
/
T
utor
ia
ls
.
S
i
mult
a
n
e
ously
,
N
e
ur
IPS CDM
X
2025
(
N
ov
30–
D
ec
2)
o
ffe
r
ed
s
e
l
ec
t
ed
wor
k
s
h
ops
a
n
d
tutor
ia
ls
,
e
sp
ecia
lly
t
a
r
ge
t
i
n
g
L
a
t
i
n
A
m
e
r
ica
n
top
ic
s
.
F
or
e
x
a
mpl
e
,
on
N
ov
30
wor
k
s
h
ops
i
n
c
lu
ded
“
V
i
s
i
on
L
a
n
g
u
age
M
o
de
ls
:
C
ha
ll
e
n
ge
s
o
f
R
ea
l
D
e
ploym
e
nt
”
a
n
d
“
R
o
b
ust
E
m
b
o
died
R
o
b
ot
ic
s
”
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
e
mp
ha
s
i
z
i
n
g
ha
r
d
w
a
r
e
c
onstr
ai
nts
a
n
d
lo
ca
l
n
eed
s
.
A
not
he
r
ke
y
wor
k
s
h
op
w
a
s
“
AI
i
n
M
e
x
ic
o
:
M
ai
n
A
ppl
ica
t
i
ons
”,
high
l
igh
t
i
n
g
r
egi
on
a
l
AI
ad
v
a
n
ce
m
e
nts
a
n
d
cha
ll
e
n
ge
s
i
n
L
a
t
i
n
A
m
e
r
ica
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
s
ec
on
d
da
y
(
D
ec
1)
had
wor
k
s
h
ops
l
ike
“
LLM P
e
rson
a
M
o
de
l
i
n
g
”
(
add
r
e
ss
i
n
g
h
ow
to
s
i
mul
a
t
e
p
e
rson
a
s
w
i
t
h
LLM
s
s
afe
ly
)
a
n
d
“
C
e
nt
e
r
i
n
g
L
ow
-
R
e
sour
ce
L
a
n
g
u
age
s
”
(
e
nsur
i
n
g
AI
mo
de
ls
s
e
rv
e
un
de
r
-
r
e
pr
e
s
e
nt
ed
l
a
n
g
u
age
s
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
utor
ia
ls
on
D
ec
2
c
ov
e
r
ed
t
ech
n
ica
l
f
ronts
,
e
.
g
.
“
E
fficie
nt
T
r
a
ns
f
orm
e
rs
”
(
prun
i
n
g
/
a
tt
e
nt
i
on
m
e
t
h
o
d
s
)
a
n
d
“
G
e
osp
a
t
ia
l
F
oun
da
t
i
on
M
o
de
ls
”
a
ppl
ied
to
E
a
rt
h
o
b
s
e
rv
a
t
i
on
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
bif
ur
ca
t
ed
pro
g
r
a
m
m
ea
nt
t
ha
t
N
e
ur
IPS
r
e
turn
ed
to
i
ts
wor
k
s
h
op
roots
(
o
f
t
e
n
i
n
s
ki
lo
dge
s
i
n
ea
rly
y
ea
rs
)
e
v
e
n
a
s
i
t
s
ca
l
ed
up
.
T
he
M
e
x
ic
o
C
i
ty
s
i
t
e
effec
t
i
v
e
ly
d
ou
b
l
ed
ca
p
aci
ty
f
or
wor
k
s
h
ops
/
tutor
ia
ls
a
n
d
b
rou
gh
t
N
e
ur
IPS
to
a
n
e
w
a
u
die
n
ce
.
I
t
a
lso
i
nt
eg
r
a
t
ed
so
cia
l
e
n
gage
m
e
nt
(
a
s
ide
f
rom
t
ech
n
ica
l
)
w
i
t
h
e
v
e
nts
l
ike
a
so
cial
s
CDM
X
mee
tup
a
n
d
outr
each
s
e
ss
i
ons
.
5.3
I
n
d
ustry
,
S
ponsors
hi
p
,
a
n
d
E
xpo
N
e
ur
IPS
2025ʼ
s
i
n
d
ustry
i
nvolv
e
m
e
nt
w
a
s
e
xtr
a
or
di
n
a
ry
.
E
v
e
ry
y
ea
r
,
t
he
c
on
fe
r
e
n
ce
high
l
igh
ts
b
r
a
n
d
s
–
“
D
ia
mon
d
S
ponsor
”,
“
G
ol
d
S
ponsor
”,
e
t
c
.
–
a
n
d
t
he
s
e
c
omp
a
n
ie
s
run
b
oot
h
s
a
n
d
r
ec
ru
i
t
.
G
oo
g
l
e
(
D
ia
mon
d
S
ponsor
)
r
a
n
a
prom
i
n
e
nt
b
oot
h
w
i
t
h
de
mos
(
m
aki
n
g
i
ts
n
e
w
“
MOMEN
TU
M
”
age
nt
,
e
t
c
.)
a
n
d
mult
i
pl
e
Q
&
A
ki
os
k
s
(
r
e
s
ea
r
ch
.
g
oo
g
l
e
).
M
ic
roso
f
t
a
lso
st
affed
b
oot
h
s
a
n
d
s
e
nt
s
e
n
i
or
r
e
s
ea
r
che
rs
to
gi
v
e
t
a
l
k
s
.
N
um
e
rous
c
orpor
a
t
e
sponsors
(
N
v
idia
,
M
e
t
a
,
IBM
,
e
t
c
.)
fi
ll
ed
t
he
E
xpo
ha
ll
.
I
n
d
ustry
p
a
rt
ici
p
a
t
i
on
a
lso
s
h
ows
i
n
t
he
c
ont
e
nt
.
M
a
ny
wor
k
s
h
ops
w
e
r
e
c
o
-
or
ga
n
i
z
ed
b
y
c
omp
a
n
ie
s
(
e
.
g
.
t
he
A
ge
nts
S
afe
ty
c
o
-
or
ga
n
i
z
ed
b
y
CAIS
/
AISI
l
ed
b
y
i
n
d
ustry
v
e
t
e
r
a
ns
)
a
n
d
a
l
a
r
ge
f
r
ac
t
i
on
o
f
acce
pt
ed
p
a
p
e
rs
had
c
orpor
a
t
e
a
ut
h
ors
(
e
sp
ecia
lly
f
rom
G
oo
g
l
e
,
M
e
t
a
,
M
ic
roso
f
t
,
a
n
d
high
-
pro
fi
l
e
AI
l
ab
s
l
ike
O
p
e
n
AI
a
n
d
D
ee
p
M
i
n
d
).
T
he
G
oo
gle
bl
o
g
e
xpl
ici
tly
m
e
nt
i
on
ed
t
he
y
had
175
acce
pt
ed
p
a
p
e
rs
a
n
d
sponsor
ed
mult
i
pl
e
wor
k
s
h
ops
a
n
d
m
e
ntor
i
n
g
e
v
e
nts
(
r
e
s
ea
r
ch
.
g
oo
g
l
e
).
M
edia
c
omm
e
nt
a
ry
(
e
.
g
.
F
ortun
e
(
[3]
f
ortun
e
.
c
om
),
R
e
ut
e
rs
)
c
ons
i
st
e
ntly
not
e
t
ha
t
N
e
ur
IPS
ʼ
s
p
age
s
a
r
e
c
row
ded
w
i
t
h
r
ec
ru
i
t
e
rs
a
n
d
st
a
rtups
:
t
he
2023
r
e
port
i
n
g
de
s
c
r
ibed
N
e
ur
IPS
a
s
a
“
r
ec
ru
i
t
i
n
g
e
xtr
a
v
aga
nz
a
,”
a
n
d
w
hi
l
e
2025ʼ
s
ton
e
w
a
s
l
e
ss
f
r
e
nzy
,
t
he
tr
e
n
d
p
e
rs
i
sts
.
M
a
ny
a
tt
e
n
dee
s
(
e
sp
ecia
lly
stu
de
nts
a
n
d
post
d
o
c
s
)
us
e
N
e
ur
IPS
to
m
ee
t
pot
e
nt
ia
l
e
mploy
e
rs
.
I
n
fac
t
,
t
hi
s
i
n
d
ustry
sur
ge
i
s
on
e
r
ea
son
t
he
c
on
fe
r
e
n
ce
ha
ll
had
to
e
xp
a
n
d
–
not
only
to
h
ous
e
mor
e
t
ech
n
ica
l
t
a
l
k
s
,
b
ut
to
acc
ommo
da
t
e
c
omp
a
n
ie
s
ʼ
sw
ag
st
a
n
d
s
,
s
ide
-
e
v
e
nts
,
a
n
d
so
cia
l
r
ec
ru
i
t
i
n
g
ga
t
he
r
i
n
g
s
.
T
ech
gia
nts
(
G
oo
g
l
e
,
M
ic
roso
f
t
,
M
e
t
a
/
AI
)
w
e
r
e
high
ly
v
i
s
ib
l
e
,
a
n
d
e
v
e
n
ba
n
k
s
a
n
d
p
ha
rm
ace
ut
ica
l
fi
rms
had
pr
e
s
e
n
ce
.
5.4
N
eighb
or
h
oo
d
S
igh
t
i
n
g
s
a
n
d
C
on
fe
r
e
n
ce
L
ife
B
e
yon
d
f
orm
a
l
s
e
ss
i
ons
,
m
a
ny
not
e
s
o
f
“
c
on
fe
r
e
n
ce
c
ultur
e
”
e
m
e
r
ged
.
F
or
e
x
a
mpl
e
,
i
t
i
n
e
r
a
nt
m
i
n
i
-
e
v
e
nts
popp
ed
up
(
o
f
t
e
n
uno
fficia
lly
).
O
n
e
must
a
lw
a
ys
be
ca
ut
i
ous
ab
out
“
uno
fficia
l
”
ga
t
he
r
i
n
g
s
(
t
he
o
fficia
l
s
i
t
e
w
a
rns
t
ha
t
only
e
v
e
nts
l
i
st
ed
on
n
e
ur
i
ps
.
cc
a
r
e
e
n
d
ors
ed
(
n
e
ur
i
ps
.
cc
)),
b
ut
a
n
ecd
ot
a
l
r
e
ports
m
e
nt
i
on
t
hi
n
g
s
l
ike
st
a
rtup
p
i
t
ch
n
igh
ts
,
e
m
e
r
ge
nt
lun
ch
t
i
m
e
s
e
m
i
n
a
rs
,
a
n
d
ad
-
h
o
c
c
o
di
n
g
spr
i
nts
.
G
i
v
e
n
tr
a
v
e
l
f
rom
di
v
e
rs
e
fie
l
d
s
,
som
e
a
tt
e
n
dee
s
f
orm
ed
i
nt
e
r
e
st
g
roups
(
e
.
g
.
b
y
c
ountry
or
sp
ecia
l
i
z
a
t
i
on
)
t
ha
t
m
e
t
i
n
f
orm
a
lly
.
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 10 of 16
O
v
e
r
a
ll
,
N
e
ur
IPS
2025
w
a
s
e
x
ha
ust
i
n
g
b
y
m
a
ny
acc
ounts
–
t
he
pro
g
r
a
m
w
a
s
ja
m
-
p
acked
f
rom
ea
rly
morn
i
n
g
to
l
a
t
e
e
v
e
n
i
n
g
,
a
n
d
mult
i
pl
e
p
a
r
a
ll
e
l
tr
ack
s
m
ea
nt
tou
gh
s
ched
ul
i
n
g
ch
o
ice
s
.
B
ut
i
t
a
lso
e
n
ab
l
ed
s
e
r
e
n
di
p
i
ty
:
hea
r
i
n
g
som
e
t
hi
n
g
outs
ide
on
e
ʼ
s
n
iche
,
m
ee
t
i
n
g
i
nt
e
rn
a
t
i
on
a
l
c
oll
ab
or
a
tors
,
or
s
i
mply
e
n
j
oy
i
n
g
t
he
ci
ty
.
T
he
i
n
c
lus
i
on
o
f
M
e
x
ic
o
C
i
ty
e
xt
e
n
ded
t
he
s
e
n
e
twor
k
s
i
nto
L
a
t
i
n
A
m
e
r
ica
,
poss
ib
ly
pl
a
nt
i
n
g
s
eed
s
f
or
mor
e
r
egi
on
a
l
c
on
fe
r
e
n
ce
s
i
n
f
utur
e
.
6.
D
a
t
a
a
n
d
E
v
ide
n
ce
:
T
r
e
n
d
s
A
n
a
lys
i
s
T
o
a
n
a
lyz
e
t
he
st
a
t
e
o
f
r
e
s
ea
r
ch
a
t
N
e
ur
IPS
2025,
w
e
d
r
a
w
on
qu
a
nt
i
t
a
t
i
v
e
da
t
a
a
v
ai
l
ab
l
e
t
h
rou
gh
o
fficia
l
cha
nn
e
ls
.
T
he
N
e
ur
IPS
b
lo
g
a
n
d
O
p
e
n
R
e
v
ie
w
da
t
a
prov
ide
g
l
i
mps
e
s
i
nto
su
b
m
i
ss
i
on
c
ont
e
nt
a
n
d
c
ommun
i
ty
r
e
spons
e
s
.
B
e
low
a
r
e
som
e
ke
y
fi
n
di
n
g
s
:
G
rowt
h
i
n
S
u
b
m
i
ss
i
ons
:
A
s
not
ed
,
M
ai
n
T
r
ack
su
b
m
i
ss
i
ons
r
eached
21,575
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
r
ef
l
ec
ts
c
ont
i
nu
ed
e
xpon
e
nt
ia
l
g
rowt
h
f
rom
~9,400
i
n
2020
to
13,330
i
n
2023
(
[1]
f
ortun
e
.
c
om
)
a
n
d
d
ou
b
l
i
n
g
s
i
n
ce
2024.
T
he
D
a
t
a
s
e
ts
tr
ack
s
i
m
i
l
a
rly
g
r
e
w
(
f
rom
~910
i
n
2023
to
1,820
i
n
2024,
a
n
d
1,995
i
n
2025
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)),
high
l
igh
t
i
n
g
t
he
c
ommun
i
ty
ʼ
s
i
n
c
r
ea
s
i
n
g
output
o
f
da
t
a
s
e
t
/
be
n
ch
m
a
r
k
stu
die
s
.
A
ut
h
ors
hi
p
a
n
d
D
e
mo
g
r
a
p
hic
s
:
D
e
t
ai
l
ed
de
mo
g
r
a
p
hic
s
(
ge
n
de
r
,
ge
o
g
r
a
p
h
y
)
a
r
e
not
y
e
t
pu
b
l
i
s
hed
f
or
2025,
b
ut
tr
e
n
d
s
su
gge
st
slow
di
v
e
rs
ifica
t
i
on
.
T
he
e
x
i
st
e
n
ce
o
f
7
affi
n
i
ty
e
v
e
nts
(
wom
e
n
,
L
a
t
i
nx
,
Q
u
ee
r
,
M
usl
i
ms
,
I
n
dige
nous
,
e
t
c
.)
i
n
dica
t
e
s
a
tt
e
nt
i
on
to
i
n
c
lus
i
on
.
T
he
B
r
idgi
n
g
g
r
a
nts
pro
g
r
a
m
ʼ
s
f
o
c
us
on
e
qu
i
ty
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
a
lso
s
ig
n
a
ls
eff
orts
to
b
ro
ade
n
p
a
rt
ici
p
a
t
i
on
.
H
ow
e
v
e
r
,
N
e
ur
IPS
(
l
ike
mu
ch
o
f
AI
)
st
i
ll
g
r
a
ppl
e
s
w
i
t
h
un
de
r
-
r
e
pr
e
s
e
nt
a
t
i
on
o
f
som
e
g
roups
.
P
opul
a
r
K
e
ywor
d
s
/
T
op
ic
s
:
A
n
uno
fficia
l
a
n
a
lys
i
s
o
f
acce
pt
ed
-
p
a
p
e
r
ke
ywor
d
s
(
e
.
g
.
f
rom
F
ig
ur
e
1
o
f
t
he
D
&
B
b
lo
g
(
b
lo
g
.
n
e
ur
i
ps
.
cc
))
c
on
fi
rms
t
he
d
om
i
n
a
n
ce
o
f
“
la
n
g
u
age
”
,
“
v
i
s
i
on
”
,
“
r
ei
n
f
or
ceme
nt
”
,
“
ge
n
e
r
a
t
i
v
e
”
,
“
lea
rn
i
n
g
”
,
“
unsup
e
rv
i
s
ed
”
,
e
t
c
.
I
n
t
he
D
a
t
a
s
e
ts
tr
ack
,
ke
ywor
d
s
l
ike
“
LLM
”
,
“
NLP
”
,
“
s
cie
n
ce
”
,
“
image
”
,
“
ro
b
ot
ic
s
”
w
e
r
e
pr
e
v
a
l
e
nt
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
su
gge
sts
t
he
m
ai
n
f
o
c
us
on
LLM
e
v
a
lu
a
t
i
on
,
mult
i
mo
da
l
mo
de
ls
,
a
n
d
s
cie
nt
ific
a
ppl
ica
t
i
ons
.
D
a
t
a
s
e
t
R
e
us
e
:
A
cc
or
di
n
g
to
t
he
D
&
B
surv
e
y
,
a
m
aj
or
i
ty
o
f
acce
pt
ed
p
a
p
e
rs
i
n
t
ha
t
tr
ack
(84
%
)
i
ntro
d
u
ced
b
r
a
n
d
-
n
e
w
da
t
a
c
oll
ec
t
i
ons
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
mon
g
h
ost
i
n
g
pl
a
t
f
orms
f
or
da
t
a
s
e
ts
,
w
ide
ly
us
ed
r
e
pos
i
tor
ie
s
d
om
i
n
a
t
ed
:
ov
e
r
80
%
o
f
acce
pt
ed
da
t
a
s
e
ts
w
e
r
e
on
H
u
ggi
n
g
F
ace
,
K
agg
l
e
,
D
a
t
a
v
e
rs
e
,
or
O
p
e
n
ML
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
c
on
ce
ntr
a
t
i
on
i
mpl
ie
s
a
de
fac
to
st
a
n
da
r
di
z
a
t
i
on
o
f
da
t
a
s
ha
r
i
n
g
pr
ac
t
ice
s
.
S
urv
e
y
F
eedback
:
P
ost
-
c
on
fe
r
e
n
ce
surv
e
ys
(851
a
ut
h
ors
,
155
r
e
v
ie
w
e
rs
)
w
e
r
e
c
on
d
u
c
t
ed
.
82
%
o
f
a
ut
h
ors
f
oun
d
t
he
da
t
a
h
ost
i
n
g
pro
ce
ss
smoot
h
,
t
h
ou
gh
som
e
l
a
r
ge
/
diffic
ult
da
t
a
s
e
ts
ca
us
ed
i
ssu
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
58
%
o
f
a
ut
h
ors
fe
lt
n
e
w
h
ost
i
n
g
/
m
e
t
ada
t
a
rul
e
s
l
ed
to
be
tt
e
r
r
e
v
ie
ws
,
b
ut
25
%
st
i
ll
s
a
w
room
f
or
r
e
v
ie
w
qu
a
l
i
ty
i
mprov
e
m
e
nts
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
mon
g
r
e
v
ie
w
e
rs
,
77
%
c
oul
d
ea
s
i
ly
acce
ss
da
t
a
s
e
ts
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
utom
a
t
ed
m
e
t
ada
t
a
r
e
ports
w
e
r
e
us
ef
ul
to
~70
%
o
f
r
e
v
ie
w
e
rs
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
s
e
num
be
rs
i
n
dica
t
e
a
l
a
r
ge
ly
pos
i
t
i
v
e
r
ece
pt
i
on
to
n
e
w
da
t
a
tr
ack
pol
icie
s
,
w
i
t
h
c
onstru
c
t
i
v
e
feedback
on
i
mprov
i
n
g
i
nstru
c
t
i
ons
a
n
d
tools
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
da
t
a
su
gge
sts
t
ha
t
N
e
ur
IPS
2025ʼ
s
i
n
i
t
ia
t
i
v
e
s
(
da
t
a
tr
ack
up
g
r
ade
s
,
m
e
t
ada
t
a
r
e
qu
i
r
e
m
e
nts
,
tr
a
ns
c
r
i
pt
i
on
st
a
n
da
r
d
s
l
ike
C
ro
i
ss
a
nt
)
a
r
e
mov
i
n
g
i
n
t
he
r
igh
t
di
r
ec
t
i
on
,
t
h
ou
gh
t
he
c
ommun
i
ty
i
s
st
i
ll
acc
l
i
m
a
t
i
n
g
.
T
he
tr
e
n
d
i
s
tow
a
r
d
g
r
ea
t
e
r
r
e
pro
d
u
cibi
l
i
ty
a
n
d
tr
a
nsp
a
r
e
n
c
y
(
e
.
g
.
r
e
qu
i
r
i
n
g
pu
b
l
ic
da
t
a
s
e
t
h
ost
i
n
g
)
–
a
s
ig
n
ifica
nt
c
ultur
a
l
s
hif
t
f
rom
ea
rl
ie
r
N
e
ur
IPS
y
ea
rs
.
7.
C
a
s
e
S
tu
die
s
a
n
d
E
x
a
mpl
e
s
T
o
i
llustr
a
t
e
t
he
c
ont
e
nt
a
n
d
i
mp
ac
t
o
f
N
e
ur
IPS
2025,
w
e
high
l
igh
t
two
ca
s
e
e
x
a
mpl
e
s
:
7.1
B
e
st
P
a
p
e
r
:
G
a
t
ed
A
tt
e
nt
i
on
i
n
LLM
s
G
a
t
ed
A
tt
e
nt
i
on
f
or
L
a
r
ge
L
a
n
g
u
age
M
o
del
s
:
N
on
-
li
n
ea
r
i
ty
,
S
p
a
rs
i
ty
,
a
n
d
A
tt
e
nt
i
on
-
S
i
n
k
-
F
r
ee
(
b
y
Z
iha
n
Q
i
u
e
t
a
l
.)
won
t
he
N
e
ur
IPS
2025
B
e
st
P
a
p
e
r
A
w
a
r
d
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
p
a
p
e
r
syst
e
m
a
t
ica
lly
e
v
a
lu
a
t
ed
h
ow
addi
n
g
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 11 of 16
ga
t
i
n
g
m
echa
n
i
sms
to
T
r
a
ns
f
orm
e
r
a
tt
e
nt
i
on
ca
n
i
mprov
e
LLM
p
e
r
f
orm
a
n
ce
.
T
he
y
e
xp
e
r
i
m
e
nt
ed
w
i
t
h
15
B M
o
E
mo
de
ls
a
n
d
1.7
B
de
ns
e
mo
de
ls
on
m
a
ss
i
v
e
c
orpor
a
(
up
to
3.5
T
to
ke
ns
)
a
n
d
f
oun
d
t
ha
t
a
s
i
mpl
e
s
ig
mo
id
ga
t
e
a
ppl
ied
af
t
e
r
t
he
s
ca
l
ed
-
d
ot
-
pro
d
u
c
t
a
tt
e
nt
i
on
head
s
y
ie
l
ded
c
ons
i
st
e
nt
gai
ns
.
K
e
y
fi
n
di
n
g
s
w
e
r
e
:
N
onl
i
n
ea
r
i
ty
a
n
d
S
p
a
rs
i
ty
:
T
he
ga
t
i
n
g
i
ntro
d
u
ce
s
a
n
e
xtr
a
nonl
i
n
ea
r
i
ty
a
n
d
a
llows
each
a
tt
e
nt
i
on
head
to
decide
d
yn
a
m
ica
lly
w
he
t
he
r
to
c
ontr
ib
ut
e
.
T
hi
s
a
ll
e
v
ia
t
e
s
“
a
tt
e
nt
i
on
s
i
n
k
”,
a
p
he
nom
e
non
w
he
r
e
som
e
ac
t
i
v
a
t
i
ons
bec
om
e
e
x
ce
ss
i
v
e
ly
l
a
r
ge
a
n
d
hi
n
de
r
l
ea
rn
i
n
g
.
T
he
ga
t
e
s
n
a
tur
a
lly
suppr
e
ss
ed
un
he
lp
f
ul
s
ig
n
a
ls
.
S
t
abi
l
i
ty
a
n
d
S
ca
l
abi
l
i
ty
:
M
o
de
ls
w
i
t
h
ga
t
ed
a
tt
e
nt
i
on
s
h
ow
ed
be
tt
e
r
st
abi
l
i
ty
(
tol
e
r
a
t
i
n
g
highe
r
l
ea
rn
i
n
g
r
a
t
e
s
)
a
n
d
sup
e
r
i
or
e
xtr
a
pol
a
t
i
on
to
lon
ge
r
c
ont
e
xts
t
ha
n
un
ga
t
ed
ba
s
e
l
i
n
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
E
a
s
e
o
f
A
d
opt
i
on
:
I
mport
a
ntly
,
t
hei
r
be
st
ga
t
i
n
g
de
s
ig
n
(
head
-
sp
ecific
s
ig
mo
id
)
i
s
i
mpl
e
m
e
nt
a
t
i
on
-
s
i
mpl
e
a
n
d
c
omp
a
t
ib
l
e
w
i
t
h
e
x
i
st
i
n
g
T
r
a
ns
f
orm
e
r
c
o
de
.
T
he
a
ut
h
ors
e
v
e
n
s
ha
r
ed
c
o
de
a
n
d
pr
e
tr
ai
n
ed
mo
de
ls
(
on
H
u
ggi
n
g
F
ace
)
to
faci
l
i
t
a
t
e
ad
opt
i
on
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
a
ut
h
ors
ʼ
e
xt
e
ns
i
v
e
e
xp
e
r
i
m
e
nts
(30
mo
de
l
v
a
r
ia
nts
)
a
n
d
c
l
ea
r
ab
l
a
t
i
on
m
ade
a
c
omp
e
ll
i
n
g
ca
s
e
.
T
he
B
e
st
P
a
p
e
r
C
omm
i
tt
ee
not
ed
t
ha
t
t
hi
s
wor
k
“
r
e
pr
e
s
e
nts
a
su
b
st
a
nt
ia
l
a
mount
o
f
wor
k
poss
ib
l
e
only
w
i
t
h
i
n
d
ustr
ia
l
s
ca
l
e
c
omput
i
n
g
”,
a
n
d
pr
ai
s
ed
t
he
op
e
n
r
e
l
ea
s
e
o
f
mo
de
ls
a
n
d
a
n
a
lys
i
s
i
n
a
n
e
r
a
o
f
c
los
ed
r
e
s
ea
r
ch
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
I
n
deed
,
t
he
p
a
p
e
r
ʼ
s
i
n
f
lu
e
n
ce
i
s
l
ike
ly
to
be
qu
ick
ly
fe
lt
,
a
s
i
t
prov
ide
s
a
low
-
c
ost
cha
n
ge
w
i
t
h
pot
e
nt
ia
lly
b
ro
ad
us
e
i
n
f
utur
e
LLM
a
r
chi
t
ec
tur
e
s
.
7.2
I
nv
i
t
ed
T
a
l
k
P
a
n
e
l
:
AI
i
n
S
cie
nt
ific
P
ee
r
R
e
v
ie
w
O
n
e
o
f
t
he
so
cia
l
e
v
e
nts
,
“
T
he
R
ol
e
o
f
AI
i
n
S
cie
nt
ific
P
ee
r
R
e
v
ie
w
”
(
D
ec
3,
af
t
e
rnoon
,
S
a
n
D
ieg
o
),
e
x
e
mpl
ifie
s
h
ow
N
e
ur
IPS
i
nt
eg
r
a
t
e
s
f
orw
a
r
d
-
loo
ki
n
g
deba
t
e
s
.
O
r
ga
n
i
z
ed
b
y
AI
e
t
hic
s
r
e
s
ea
r
che
rs
,
t
hi
s
e
v
e
nt
b
rou
gh
t
to
ge
t
he
r
s
cie
nt
i
sts
,
pu
b
l
i
s
he
rs
,
a
n
d
pr
ac
t
i
t
i
on
e
rs
to
di
s
c
uss
h
ow
AI
ca
n
tr
a
ns
f
orm
a
ll
st
age
s
o
f
acade
m
ic
pu
b
l
i
s
hi
n
g
.
T
op
ic
s
i
n
c
lu
ded
a
utom
a
t
ed
r
e
v
ie
w
i
n
g
tools
,
AI
-
a
ss
i
st
ed
a
ut
h
or
i
n
g
,
a
n
d
e
t
hica
l
c
ons
ide
r
a
t
i
ons
(
bia
s
i
n
r
e
v
ie
w
a
l
g
or
i
t
h
ms
,
tr
a
nsp
a
r
e
n
c
y
de
m
a
n
d
s
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
p
a
n
e
l
,
a
tt
e
n
ded
b
y
ab
out
100
p
a
rt
ici
p
a
nts
,
i
nv
e
st
iga
t
ed
qu
e
st
i
ons
l
ike
:
C
a
n
m
achi
n
e
l
ea
rn
i
n
g
f
l
ag
pro
b
l
e
m
a
t
ic
p
a
p
e
rs
or
su
gge
st
r
e
v
ie
w
e
rs
?
H
ow
d
o
w
e
pr
e
v
e
nt
AI
f
rom
chea
t
i
n
g
p
ee
r
r
e
v
ie
w
(
e
.
g
.
fab
r
ica
t
i
n
g
ci
t
a
t
i
ons
)?
T
he
di
s
c
uss
i
on
un
de
rs
c
or
ed
t
ha
t
a
s
N
e
ur
IPS
(
a
n
d
ML
r
e
s
ea
r
ch
)
g
rows
,
so
too
d
o
e
s
i
nt
e
r
e
st
i
n
“
c
los
i
n
g
t
he
loop
”
on
t
he
pu
b
l
ica
t
i
on
pro
ce
ss
.
T
hi
s
e
v
e
nt
i
s
a
m
ic
ro
c
osm
o
f
N
e
ur
IPS
ʼ
s
c
ultur
e
–
t
ech
n
ica
lly
sop
hi
st
ica
t
ed
y
e
t
so
cia
lly
c
ons
ci
ous
.
8.
I
mpl
ica
t
i
ons
a
n
d
F
utur
e
D
i
r
ec
t
i
ons
N
e
ur
IPS
2025ʼ
s
out
c
om
e
s
ha
v
e
i
mm
edia
t
e
a
n
d
lon
g
-
t
e
rm
i
mpl
ica
t
i
ons
f
or
AI
r
e
s
ea
r
ch
:
R
e
pro
d
u
cibi
l
i
ty
a
n
d
D
a
t
a
S
t
a
n
da
r
d
s
:
T
he
D
a
t
a
s
e
ts
tr
ack
e
xp
e
r
ie
n
ce
s
h
ow
ed
t
ha
t
r
e
qu
i
r
i
n
g
pu
b
l
ic
da
t
a
h
ost
i
n
g
a
n
d
m
e
t
ada
t
a
f
un
da
m
e
nt
a
lly
r
ai
s
ed
r
e
v
ie
w
tr
a
nsp
a
r
e
n
c
y
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
G
o
i
n
g
f
orw
a
r
d
,
N
e
ur
IPS
w
i
ll
l
ike
ly
t
igh
t
e
n
t
he
s
e
r
e
qu
i
r
e
m
e
nts
,
p
e
r
ha
ps
e
n
f
or
ci
n
g
c
ompl
e
t
e
R
e
spons
ib
l
e
AI
d
o
c
um
e
nt
a
t
i
on
.
F
utur
e
c
on
fe
r
e
n
ce
s
m
a
y
e
xt
e
n
d
t
hi
s
to
c
o
de
a
n
d
mo
de
l
r
e
l
ea
s
e
,
ai
m
i
n
g
f
or
f
ully
i
nt
e
rop
e
r
ab
l
e
r
e
s
ea
r
ch
a
rt
ifac
ts
.
H
ow
e
v
e
r
,
or
ga
n
i
z
e
rs
not
ed
c
urr
e
nt
ga
ps
:
a
ut
h
ors
n
eed
c
l
ea
r
e
r
g
u
ide
l
i
n
e
s
on
l
ice
ns
i
n
g
a
n
d
m
achi
n
e
-
r
eadab
l
e
de
s
c
r
i
ptors
(
b
lo
g
.
n
e
ur
i
ps
.
cc
),
a
n
d
i
n
f
r
a
stru
c
tur
e
must
be
tt
e
r
ha
n
d
l
e
m
a
ss
i
v
e
da
t
a
s
e
ts
(1+
T
B
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
A
dd
r
e
ss
i
n
g
t
he
s
e
w
i
ll
be
pr
i
or
i
ty
i
n
c
om
i
n
g
y
ea
rs
.
R
e
v
ie
w
e
r
P
ool
E
xp
a
ns
i
on
:
N
e
ur
IPS
2025
high
l
igh
t
ed
a
s
h
ort
age
o
f
top
ic
e
xp
e
rts
f
or
n
e
w
fie
l
d
s
(
LLM
s
,
RL
,
AI
f
or
s
cie
n
ce
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
he
or
ga
n
i
z
i
n
g
c
omm
i
tt
ee
r
ec
o
g
n
i
z
ed
“
l
i
m
i
t
a
t
i
ons
i
n
r
e
v
ie
w
e
r
di
v
e
rs
i
ty
a
n
d
d
om
ai
n
c
ov
e
r
age
”
a
n
d
su
gge
st
ed
r
ec
ru
i
t
i
n
g
be
yon
d
t
he
usu
a
l
ML
c
row
d
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
E
xp
ec
t
f
utur
e
eff
orts
to
d
r
a
w
i
n
st
a
t
i
st
icia
ns
,
d
om
ai
n
s
cie
nt
i
sts
,
a
n
d
i
nt
e
r
di
s
ci
pl
i
n
a
ry
r
e
s
ea
r
che
rs
a
s
r
e
v
ie
w
e
rs
.
T
he
su
cce
ss
o
f
t
he
J
ourn
a
l
tr
ack
su
gge
sts
f
urt
he
r
c
ross
-
v
e
nu
e
c
oll
ab
or
a
t
i
ons
m
a
y
a
lso
he
lp
,
b
y
a
llow
i
n
g
j
ourn
a
l
-
a
ut
h
or
ed
e
xp
e
rts
to
e
n
gage
w
i
t
h
c
on
fe
r
e
n
ce
r
e
v
ie
w
i
n
g
.
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 12 of 16
R
e
spons
ib
l
e
AI F
o
c
us
:
T
he
n
e
w
P
os
i
t
i
on
P
a
p
e
r
tr
ack
a
n
d
t
he
ton
e
o
f
i
nv
i
t
ed
l
ec
tur
e
s
i
n
dica
t
e
g
row
i
n
g
a
tt
e
nt
i
on
to
AI
ʼ
s
so
cie
t
a
l
r
a
m
ifica
t
i
ons
.
F
utur
e
N
e
ur
IPS
l
ike
ly
w
i
ll
m
ai
nt
ai
n
a
n
d
p
e
r
ha
ps
e
xp
a
n
d
f
orums
f
or
e
t
hic
s
,
pol
ic
y
,
a
n
d
p
hi
losop
h
y
.
F
or
e
x
a
mpl
e
,
p
a
n
e
ls
w
i
t
h
pol
ic
ym
ake
rs
or
non
-
t
ech
n
ica
l
st
akeh
ol
de
rs
m
igh
t
bec
om
e
c
ommon
.
T
he
b
r
idgi
n
g
g
r
a
nts
pro
g
r
a
m
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
a
n
d
wor
k
s
h
ops
on
low
-
r
e
sour
ce
or
i
n
c
lus
i
v
e
AI
hi
nt
a
t
stru
c
tur
a
l
c
omm
i
tm
e
nts
(
f
un
di
n
g
,
tr
ack
s
)
to
de
mo
c
r
a
t
i
z
e
AI
.
C
ommun
i
ty
E
n
gage
m
e
nt
:
P
ro
g
r
a
m
chai
rs
e
xpl
ici
tly
i
nv
i
t
ed
feedback
(
e
.
g
.
a
t
t
he
T
own
H
a
ll
)
to
r
efi
n
e
pro
ce
ss
e
s
(
b
lo
g
.
n
e
ur
i
ps
.
cc
).
T
hi
s
p
a
rt
ici
p
a
tory
st
a
n
ce
m
ea
ns
cha
n
ge
s
(
l
ike
t
he
LLM
pol
ic
y
)
w
i
ll
e
volv
e
w
i
t
h
c
ommun
i
ty
n
eed
s
.
I
f
2025ʼ
s
l
e
ssons
i
n
c
lu
ded
“
r
e
v
ie
w
e
r
fa
t
ig
u
e
”
f
rom
e
n
d
l
e
ss
r
eb
utt
a
ls
,
f
utur
e
y
ea
rs
m
igh
t
s
h
ort
e
n
r
e
v
ie
w
p
ha
s
e
s
or
add
m
id
-
c
y
c
l
e
r
e
v
ie
w
e
r
e
x
cha
n
ge
s
.
T
he
us
e
o
f
a
utom
a
t
ed
tools
(
l
ike
m
e
t
a
s
c
or
e
s
or
AI
-
a
ss
i
st
a
nts
f
or
r
e
v
ie
w
i
n
g
)
c
oul
d
a
lso
be
p
i
lot
ed
.
S
ca
l
e
a
n
d
V
e
nu
e
C
ons
ide
r
a
t
i
ons
:
N
e
ur
IPS
2025
st
i
ll
m
a
n
aged
p
h
ys
ica
lly
i
n
a
l
a
r
ge
c
onv
e
nt
i
on
ce
nt
e
r
,
b
ut
w
i
t
h
ba
r
e
ly
c
om
f
ort
ab
l
e
m
a
r
gi
ns
.
W
i
t
h
su
b
m
i
ss
i
ons
s
h
ow
i
n
g
no
s
ig
n
o
f
slow
i
n
g
,
f
utur
e
c
on
fe
r
e
n
ce
s
face
ca
p
aci
ty
l
i
m
i
ts
.
U
nl
ike
m
a
ny
fie
l
d
s
w
hich
opt
f
or
mult
i
pl
e
sm
a
ll
e
r
c
on
fe
r
e
n
ce
s
,
N
e
ur
IPS
ha
s
ch
os
e
n
to
r
e
m
ai
n
s
i
n
g
l
e
-
defi
n
i
t
i
v
e
a
nnu
a
l
e
v
e
nt
.
O
pt
i
ons
ahead
i
n
c
lu
de
f
urt
he
r
e
xp
a
n
di
n
g
s
a
t
e
ll
i
t
e
v
e
nu
e
s
(
e
.
g
.
N
e
ur
IP
S
E
urop
e
/
A
f
r
ica
)
or
h
y
b
r
id
mo
de
ls
.
A
lr
ead
y
,
a
E
urop
ea
n
v
e
rs
i
on
ca
ll
ed
“
EP
i
C
”
i
s
i
m
agi
n
ed
.
M
ai
nt
ai
n
i
n
g
qu
a
l
i
ty
w
hi
l
e
c
op
i
n
g
w
i
t
h
s
ca
l
e
w
i
ll
be
a
n
on
g
o
i
n
g
cha
ll
e
n
ge
.
R
e
s
ea
r
ch
D
i
r
ec
t
i
ons
:
T
he
idea
s
s
h
ow
ca
s
ed
w
i
ll
r
i
ppl
e
f
orw
a
r
d
.
T
he
su
cce
ss
o
f
ga
t
ed
a
tt
e
nt
i
on
m
a
y
i
nsp
i
r
e
mor
e
a
r
chi
t
ec
tur
a
l
tw
eak
s
i
n
LLM
s
.
T
he
c
r
i
t
ica
l
fi
n
di
n
g
s
on
RL
su
gge
st
n
e
w
r
e
s
ea
r
ch
i
nto
c
ont
i
nu
a
l
a
n
d
mult
i
-
turn
tr
ai
n
i
n
g
.
T
he
prom
i
n
e
n
ce
o
f
age
nt
ic
AI
(“
AI
a
ss
i
st
a
nts
w
i
t
h
tools
”)
po
i
nts
to
l
ike
ly
g
rowt
h
i
n
be
n
ch
m
a
r
k
de
v
e
lopm
e
nt
f
or
age
nts
(
a
s
som
e
wor
k
s
h
ops
i
n
dica
t
ed
).
I
n
s
h
ort
,
c
on
fe
r
e
n
ce
s
l
ike
N
e
ur
IPS
o
f
t
e
n
s
ig
n
a
l
r
e
s
ea
r
ch
f
o
ci
;
2025
s
ig
n
a
ls
a
dee
p
di
v
e
i
nto
LLM
i
nt
e
rn
a
ls
,
a
r
eck
on
i
n
g
on
AI
i
mp
ac
t
,
a
n
d
c
ont
i
nu
ed
dee
p
e
n
i
n
g
o
f
ML
t
he
ory
.
9.
C
on
c
lus
i
on
N
e
ur
IPS
2025
e
p
i
tom
i
z
ed
a
fie
l
d
a
t
a
c
rossro
ad
s
.
I
t
w
a
s
b
ot
h
t
he
c
ulm
i
n
a
t
i
on
o
f
m
achi
n
e
l
ea
rn
i
n
g
ʼ
s
e
xplos
i
v
e
g
rowt
h
,
a
n
d
a
s
e
l
f
-
e
x
a
m
i
n
a
t
i
on
o
f
w
he
r
e
t
ha
t
g
rowt
h
l
ead
s
.
T
he
c
on
fe
r
e
n
ce
c
om
bi
n
ed
m
a
ss
i
v
e
s
ca
l
e
(
ov
e
r
21
K
su
b
m
i
ss
i
ons
,
t
h
ous
a
n
d
s
o
f
a
tt
e
n
dee
s
)
w
i
t
h
n
e
w
m
echa
n
i
sms
f
or
r
ef
l
ec
t
i
on
(
P
os
i
t
i
on
P
a
p
e
rs
,
e
t
hic
s
wor
k
s
h
ops
,
LLM
pol
icie
s
).
T
he
t
ech
n
ica
l
pro
g
r
a
m
s
h
ow
ca
s
ed
b
ot
h
t
he
pow
e
r
o
f
mo
de
rn
AI
(
LLM
s
,
RL
,
f
oun
da
t
i
on
mo
de
ls
)
a
n
d
i
ts
l
i
m
i
t
a
t
i
ons
(
s
h
ort
c
om
i
n
g
s
i
n
r
ea
son
i
n
g
,
da
t
a
bia
s
e
s
).
K
e
ynot
e
s
a
n
d
a
w
a
r
d
s
h
onor
ed
b
ot
h
c
utt
i
n
g
-
edge
i
nnov
a
t
i
on
a
n
d
f
oun
da
t
i
on
a
l
idea
s
.
F
rom
a
n
or
ga
n
i
z
a
t
i
on
a
l
p
e
rsp
ec
t
i
v
e
,
N
e
ur
IPS
2025
ad
m
i
r
ab
ly
ha
n
d
l
ed
unpr
ecede
nt
ed
lo
gi
st
ica
l
de
m
a
n
d
s
.
R
e
v
ie
w
e
rs
a
n
d
chai
rs
m
a
n
aged
tw
e
nty
t
h
ous
a
n
d
su
b
m
i
ss
i
ons
;
pol
icie
s
w
e
r
e
r
efi
n
ed
to
m
ai
nt
ai
n
i
nt
eg
r
i
ty
i
n
a
n
e
r
a
o
f
ge
n
e
r
a
t
i
v
e
tools
a
n
d
s
ec
ur
i
ty
c
on
ce
rns
.
T
he
di
v
e
rs
e
a
rr
a
y
o
f
e
v
e
nts
–
f
rom
so
cia
l
c
r
ea
t
i
v
i
ty
ja
ms
to
d
om
ai
n
-
sp
ecific
wor
k
s
h
ops
–
i
llustr
a
t
ed
a
v
ib
r
a
nt
a
n
d
c
ons
cie
nt
i
ous
c
ommun
i
ty
.
L
oo
ki
n
g
f
orw
a
r
d
,
N
e
ur
IPS
ha
s
s
e
t
a
n
e
w
be
n
ch
m
a
r
k
.
I
ts
e
xp
a
n
ded
tr
ack
s
a
n
d
pro
g
r
a
ms
m
a
y
s
e
rv
e
a
s
a
mo
de
l
f
or
ot
he
r
c
on
fe
r
e
n
ce
s
wr
e
stl
i
n
g
w
i
t
h
s
i
m
i
l
a
r
g
rowt
h
.
T
he
r
ig
orous
d
o
c
um
e
nt
a
t
i
on
pr
ac
t
ice
s
(
da
t
a
s
e
t
h
ost
i
n
g
,
LLM
us
age
pol
ic
y
)
a
n
d
c
ommun
i
ty
-
b
u
i
l
di
n
g
eff
orts
(
affi
n
i
ty
e
v
e
nts
,
b
r
idgi
n
g
g
r
a
nts
)
s
h
ow
a
c
on
fe
r
e
n
ce
w
i
ll
i
n
g
to
e
volv
e
i
n
e
mp
i
r
ica
l
r
ig
or
a
n
d
i
n
c
lus
i
v
i
ty
.
A
s
AI
c
ont
i
nu
e
s
to
e
volv
e
,
N
e
ur
IPS
w
i
ll
l
ike
ly
r
e
m
ai
n
i
ts
s
h
ow
ca
s
e
a
n
d
t
hi
n
k
-
t
a
n
k
–
a
n
d
2025ʼ
s
r
ich
di
s
c
uss
i
ons
a
n
d
fi
n
di
n
g
s
w
i
ll
r
e
v
e
r
be
r
a
t
e
i
n
t
he
fie
l
d
ʼ
s
n
e
xt
m
i
l
e
ston
e
s
.
I
n
sum
,
N
e
ur
IPS
2025
w
a
s
a
m
i
l
e
ston
e
e
v
e
nt
,
not
ab
l
e
f
or
i
ts
s
ca
l
e
,
i
ts
su
b
st
a
nt
i
v
e
AI
r
e
s
ea
r
ch
,
a
n
d
i
ts
ada
pt
i
v
e
pol
icie
s
.
T
hi
s
r
e
port
ha
s
de
t
ai
l
ed
t
he
k
nown
fac
ts
a
n
d
high
l
igh
ts
o
f
t
ha
t
c
on
fe
r
e
n
ce
,
ci
t
i
n
g
o
fficia
l
sour
ce
s
a
n
d
a
n
a
lys
e
s
t
h
rou
gh
out
.
A
ll
c
l
ai
ms
he
r
ei
n
a
r
e
support
ed
b
y
t
he
N
e
ur
IPS
b
lo
g
,
G
oo
g
l
e
a
n
d
M
ic
roso
f
t
pu
b
l
ica
t
i
ons
,
n
e
ws
a
rt
ic
l
e
s
,
a
n
d
t
he
N
e
ur
IPS
pro
g
r
a
m
m
a
t
e
r
ia
ls
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
r
e
s
ea
r
ch
.
g
oo
g
l
e
)
(
b
lo
g
.
n
e
ur
i
ps
.
cc
)
(
st
a
t
ic
.
h
lt
.
b
m
e
.
h
u
).
T
he
f
ull
i
mp
ac
t
o
f
N
e
ur
IPS
2025
w
i
ll
un
f
ol
d
i
n
t
he
c
om
i
n
g
y
ea
rs
a
s
r
e
s
ea
r
ch
b
u
i
l
d
s
on
i
ts
di
s
c
uss
i
ons
,
b
ut
f
or
now
i
t
st
a
n
d
s
a
s
a
c
ompr
ehe
ns
i
v
e
sn
a
ps
h
ot
o
f
AI
i
n
l
a
t
e
2025.
E
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 13 of 16
E
xt
e
rn
a
l
S
our
ce
s
[1]
h
ttps
://
f
ortun
e
.
c
om
/2023/12/19/
n
e
ur
i
ps
-
ai
-
t
a
l
e
nt
-
r
ec
ru
i
t
i
n
g
-
f
r
e
nzy
/
#
:~:
b
uzzy
...
[2]
h
ttps
://
www
.
r
e
ut
e
rs
.
c
om
/
t
ech
nolo
g
y
/
a
rt
ificia
l
-
i
nt
e
ll
ige
n
ce
/
w
he
n
-
ai
-
v
ie
s
-
w
i
t
h
-
t
a
ylor
-
sw
if
t
-
h
ot
-
t
icke
t
-
town
-2024-12-16/
#
:~:
T
he
%
2...
[3]
h
ttps
://
f
ortun
e
.
c
om
/2023/12/19/
n
e
ur
i
ps
-
ai
-
t
a
l
e
nt
-
r
ec
ru
i
t
i
n
g
-
f
r
e
nzy
/
#
:~:
B
ig
%
2...
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 14 of 16
I
ntu
i
t
i
on
L
ab
s
-
I
n
d
ustry
L
eade
rs
hi
p
&
S
e
rv
ice
s
N
ort
h
A
m
e
r
ica
'
s
#
1
AI S
o
f
tw
a
r
e
D
e
v
e
lopm
e
nt
F
i
rm
f
or
P
ha
rm
ace
ut
ica
l
&
B
i
ot
ech
:
I
ntu
i
t
i
on
L
ab
s
l
ead
s
t
he
U
S
m
a
r
ke
t
i
n
c
ustom
AI
so
f
tw
a
r
e
de
v
e
lopm
e
nt
a
n
d
p
ha
rm
a
i
mpl
e
m
e
nt
a
t
i
ons
w
i
t
h
prov
e
n
r
e
sults
ac
ross
pu
b
l
ic
bi
ot
ech
a
n
d
p
ha
rm
ace
ut
ica
l
c
omp
a
n
ie
s
.
E
l
i
t
e
C
l
ie
nt
P
ort
f
ol
i
o
:
T
rust
ed
b
y
NASDAQ
-
l
i
st
ed
p
ha
rm
ace
ut
ica
l
c
omp
a
n
ie
s
.
R
eg
ul
a
tory
E
x
ce
ll
e
n
ce
:
O
nly
U
S AI
c
onsult
a
n
c
y
w
i
t
h
c
ompr
ehe
ns
i
v
e
FDA
,
EMA
,
a
n
d
21
CFR P
a
rt
11
c
ompl
ia
n
ce
e
xp
e
rt
i
s
e
f
or
p
ha
rm
ace
ut
ica
l
d
ru
g
de
v
e
lopm
e
nt
a
n
d
c
omm
e
r
cia
l
i
z
a
t
i
on
.
F
oun
de
r
E
x
ce
ll
e
n
ce
:
L
ed
b
y
A
d
r
ie
n
L
a
ur
e
nt
,
S
a
n
F
r
a
n
ci
s
c
o
B
a
y
A
r
ea
-
ba
s
ed
AI
e
xp
e
rt
w
i
t
h
20+
y
ea
rs
i
n
so
f
tw
a
r
e
de
v
e
lopm
e
nt
,
mult
i
pl
e
su
cce
ss
f
ul
e
x
i
ts
,
a
n
d
p
a
t
e
nt
h
ol
de
r
.
R
ec
o
g
n
i
z
ed
a
s
on
e
o
f
t
he
top
AI
e
xp
e
rts
i
n
t
he
U
SA
.
C
ustom
AI S
o
f
tw
a
r
e
D
e
v
e
lopm
e
nt
:
B
u
i
l
d
t
ai
lor
ed
p
ha
rm
ace
ut
ica
l
AI
a
ppl
ica
t
i
ons
,
c
ustom
CRM
s
,
cha
t
b
ots
,
a
n
d
ERP
syst
e
ms
w
i
t
h
ad
v
a
n
ced
a
n
a
lyt
ic
s
a
n
d
r
eg
ul
a
tory
c
ompl
ia
n
ce
ca
p
abi
l
i
t
ie
s
.
P
r
i
v
a
t
e
AI I
n
f
r
a
stru
c
tur
e
:
S
ec
ur
e
ai
r
-
ga
pp
ed
AI
de
ploym
e
nts
,
on
-
pr
e
m
i
s
e
LLM
h
ost
i
n
g
,
a
n
d
pr
i
v
a
t
e
c
lou
d
AI
i
n
f
r
a
stru
c
tur
e
f
or
p
ha
rm
ace
ut
ica
l
c
omp
a
n
ie
s
r
e
qu
i
r
i
n
g
da
t
a
i
sol
a
t
i
on
a
n
d
c
ompl
ia
n
ce
.
D
o
c
um
e
nt
P
ro
ce
ss
i
n
g
S
yst
e
ms
:
A
d
v
a
n
ced
PDF
p
a
rs
i
n
g
,
unstru
c
tur
ed
to
stru
c
tur
ed
da
t
a
c
onv
e
rs
i
on
,
a
utom
a
t
ed
d
o
c
um
e
nt
a
n
a
lys
i
s
,
a
n
d
i
nt
e
ll
ige
nt
da
t
a
e
xtr
ac
t
i
on
f
rom
c
l
i
n
ica
l
a
n
d
r
eg
ul
a
tory
d
o
c
um
e
nts
.
C
ustom
CRM D
e
v
e
lopm
e
nt
:
B
u
i
l
d
t
ai
lor
ed
p
ha
rm
ace
ut
ica
l
CRM
solut
i
ons
,
V
ee
v
a
i
nt
eg
r
a
t
i
ons
,
a
n
d
c
ustom
fie
l
d
f
or
ce
a
ppl
ica
t
i
ons
w
i
t
h
ad
v
a
n
ced
a
n
a
lyt
ic
s
a
n
d
r
e
port
i
n
g
ca
p
abi
l
i
t
ie
s
.
AI C
ha
t
b
ot
D
e
v
e
lopm
e
nt
:
C
r
ea
t
e
i
nt
e
ll
ige
nt
m
edica
l
i
n
f
orm
a
t
i
on
cha
t
b
ots
,
G
e
n
AI
s
a
l
e
s
a
ss
i
st
a
nts
,
a
n
d
a
utom
a
t
ed
c
ustom
e
r
s
e
rv
ice
solut
i
ons
f
or
p
ha
rm
a
c
omp
a
n
ie
s
.
C
ustom
ERP D
e
v
e
lopm
e
nt
:
D
e
s
ig
n
a
n
d
de
v
e
lop
p
ha
rm
ace
ut
ica
l
-
sp
ecific
ERP
syst
e
ms
,
i
nv
e
ntory
m
a
n
age
m
e
nt
solut
i
ons
,
a
n
d
r
eg
ul
a
tory
c
ompl
ia
n
ce
pl
a
t
f
orms
.
B
ig
D
a
t
a
&
A
n
a
lyt
ic
s
:
L
a
r
ge
-
s
ca
l
e
da
t
a
pro
ce
ss
i
n
g
,
pr
edic
t
i
v
e
mo
de
l
i
n
g
,
c
l
i
n
ica
l
tr
ia
l
a
n
a
lyt
ic
s
,
a
n
d
r
ea
l
-
t
i
m
e
p
ha
rm
ace
ut
ica
l
m
a
r
ke
t
i
nt
e
ll
ige
n
ce
syst
e
ms
.
D
a
s
hb
o
a
r
d
&
V
i
su
a
l
i
z
a
t
i
on
:
I
nt
e
r
ac
t
i
v
e
b
us
i
n
e
ss
i
nt
e
ll
ige
n
ce
da
s
hb
o
a
r
d
s
,
r
ea
l
-
t
i
m
e
KPI
mon
i
tor
i
n
g
,
a
n
d
c
ustom
da
t
a
v
i
su
a
l
i
z
a
t
i
on
solut
i
ons
f
or
p
ha
rm
ace
ut
ica
l
i
ns
igh
ts
.
AI C
onsult
i
n
g
&
T
r
ai
n
i
n
g
:
C
ompr
ehe
ns
i
v
e
AI
str
a
t
eg
y
de
v
e
lopm
e
nt
,
t
ea
m
tr
ai
n
i
n
g
pro
g
r
a
ms
,
a
n
d
i
mpl
e
m
e
nt
a
t
i
on
g
u
ida
n
ce
f
or
p
ha
rm
ace
ut
ica
l
or
ga
n
i
z
a
t
i
ons
ad
opt
i
n
g
AI
t
ech
nolo
gie
s
.
C
ont
ac
t
f
oun
de
r
A
d
r
ie
n
L
a
ur
e
nt
a
n
d
t
ea
m
a
t
h
ttps
://
i
ntu
i
t
i
onl
ab
s
.
ai
/
c
ont
ac
t
f
or
a
c
onsult
a
t
i
on
.
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 15 of 16
DISCLAIMER
T
he
i
n
f
orm
a
t
i
on
c
ont
ai
n
ed
i
n
t
hi
s
d
o
c
um
e
nt
i
s
prov
ided
f
or
ed
u
ca
t
i
on
a
l
a
n
d
i
n
f
orm
a
t
i
on
a
l
purpos
e
s
only
.
W
e
m
ake
no
r
e
pr
e
s
e
nt
a
t
i
ons
or
w
a
rr
a
nt
ie
s
o
f
a
ny
ki
n
d
,
e
xpr
e
ss
or
i
mpl
ied
,
ab
out
t
he
c
ompl
e
t
e
n
e
ss
,
acc
ur
ac
y
,
r
e
l
iabi
l
i
ty
,
su
i
t
abi
l
i
ty
,
or
a
v
ai
l
abi
l
i
ty
o
f
t
he
i
n
f
orm
a
t
i
on
c
ont
ai
n
ed
he
r
ei
n
.
A
ny
r
e
l
ia
n
ce
you
pl
ace
on
su
ch
i
n
f
orm
a
t
i
on
i
s
str
ic
tly
a
t
your
own
r
i
s
k
.
I
n
no
e
v
e
nt
w
i
ll
I
ntu
i
t
i
on
L
ab
s
.
ai
or
i
ts
r
e
pr
e
s
e
nt
a
t
i
v
e
s
be
l
iab
l
e
f
or
a
ny
loss
or
da
m
age
i
n
c
lu
di
n
g
w
i
t
h
out
l
i
m
i
t
a
t
i
on
,
i
n
di
r
ec
t
or
c
ons
e
qu
e
nt
ia
l
loss
or
da
m
age
,
or
a
ny
loss
or
da
m
age
w
ha
tso
e
v
e
r
a
r
i
s
i
n
g
f
rom
t
he
us
e
o
f
i
n
f
orm
a
t
i
on
pr
e
s
e
nt
ed
i
n
t
hi
s
d
o
c
um
e
nt
.
T
hi
s
d
o
c
um
e
nt
m
a
y
c
ont
ai
n
c
ont
e
nt
ge
n
e
r
a
t
ed
w
i
t
h
t
he
a
ss
i
st
a
n
ce
o
f
a
rt
ificia
l
i
nt
e
ll
ige
n
ce
t
ech
nolo
gie
s
.
AI
-
ge
n
e
r
a
t
ed
c
ont
e
nt
m
a
y
c
ont
ai
n
e
rrors
,
om
i
ss
i
ons
,
or
i
n
acc
ur
acie
s
.
R
eade
rs
a
r
e
ad
v
i
s
ed
to
i
n
de
p
e
n
de
ntly
v
e
r
if
y
a
ny
c
r
i
t
ica
l
i
n
f
orm
a
t
i
on
bef
or
e
ac
t
i
n
g
upon
i
t
.
A
ll
pro
d
u
c
t
n
a
m
e
s
,
lo
g
os
,
b
r
a
n
d
s
,
tr
ade
m
a
r
k
s
,
a
n
d
r
egi
st
e
r
ed
tr
ade
m
a
r
k
s
m
e
nt
i
on
ed
i
n
t
hi
s
d
o
c
um
e
nt
a
r
e
t
he
prop
e
rty
o
f
t
hei
r
r
e
sp
ec
t
i
v
e
own
e
rs
.
A
ll
c
omp
a
ny
,
pro
d
u
c
t
,
a
n
d
s
e
rv
ice
n
a
m
e
s
us
ed
i
n
t
hi
s
d
o
c
um
e
nt
a
r
e
f
or
ide
nt
ifica
t
i
on
purpos
e
s
only
.
U
s
e
o
f
t
he
s
e
n
a
m
e
s
,
lo
g
os
,
tr
ade
m
a
r
k
s
,
a
n
d
b
r
a
n
d
s
d
o
e
s
not
i
mply
e
n
d
ors
e
m
e
nt
b
y
t
he
r
e
sp
ec
t
i
v
e
tr
ade
m
a
r
k
h
ol
de
rs
.
I
ntu
i
t
i
on
L
ab
s
.
ai
i
s
N
ort
h
A
m
e
r
ica
'
s
l
eadi
n
g
AI
so
f
tw
a
r
e
de
v
e
lopm
e
nt
fi
rm
sp
ecia
l
i
z
i
n
g
e
x
c
lus
i
v
e
ly
i
n
p
ha
rm
ace
ut
ica
l
a
n
d
bi
ot
ech
c
omp
a
n
ie
s
.
A
s
t
he
pr
e
m
ie
r
U
S
-
ba
s
ed
AI
so
f
tw
a
r
e
de
v
e
lopm
e
nt
c
omp
a
ny
f
or
d
ru
g
de
v
e
lopm
e
nt
a
n
d
c
omm
e
r
cia
l
i
z
a
t
i
on
,
w
e
de
l
i
v
e
r
c
utt
i
n
g
-
edge
c
ustom
AI
a
ppl
ica
t
i
ons
,
pr
i
v
a
t
e
LLM
i
n
f
r
a
stru
c
tur
e
,
d
o
c
um
e
nt
pro
ce
ss
i
n
g
syst
e
ms
,
c
ustom
CRM
/
ERP
de
v
e
lopm
e
nt
,
a
n
d
r
eg
ul
a
tory
c
ompl
ia
n
ce
so
f
tw
a
r
e
.
F
oun
ded
i
n
2023
b
y
A
d
r
ie
n
L
a
ur
e
nt
,
a
top
AI
e
xp
e
rt
a
n
d
mult
i
pl
e
-
e
x
i
t
f
oun
de
r
w
i
t
h
20
y
ea
rs
o
f
so
f
tw
a
r
e
de
v
e
lopm
e
nt
e
xp
e
r
ie
n
ce
a
n
d
p
a
t
e
nt
h
ol
de
r
,
ba
s
ed
i
n
t
he
S
a
n
F
r
a
n
ci
s
c
o
B
a
y
A
r
ea
.
T
hi
s
d
o
c
um
e
nt
d
o
e
s
not
c
onst
i
tut
e
pro
fe
ss
i
on
a
l
or
l
ega
l
ad
v
ice
.
F
or
sp
ecific
g
u
ida
n
ce
r
e
l
a
t
ed
to
your
b
us
i
n
e
ss
n
eed
s
,
pl
ea
s
e
c
onsult
w
i
t
h
a
ppropr
ia
t
e
qu
a
l
ified
pro
fe
ss
i
on
a
ls
.
©
2025
I
ntu
i
t
i
on
L
ab
s
.
ai
.
A
ll
r
igh
ts
r
e
s
e
rv
ed
.
IntuitionLabs - Custom AI Software Development
from the leading AI expert Adrien Laurent NeurIPS 2025: A Guide to Key Papers, Trends & Stats
© 2025 IntuitionLabs.ai - North America's Leading AI Software Development Firm for Pharmaceutical & Biotech. All rights reserved. Page 16 of 16