Apperently it’s more efficient to do it other way around, to compile programs into transformers, which are then useful as refecene and ground truth when analyzing “real” transformers.
See usage of TRACR in “Towards Automated Circuit Discovery for Mechanistic Interpretability” https://arxiv.org/pdf/2304.14997, for example.
Current theme: default
Less Wrong (text)
Less Wrong (link)
Arrow keys: Next/previous image
Escape or click: Hide zoomed image
Space bar: Reset image size & position
Scroll to zoom in/out
(When zoomed in, drag to pan; double-click to close)
Keys shown in yellow (e.g., ]) are accesskeys, and require a browser-specific modifier key (or keys).
]
Keys shown in grey (e.g., ?) do not require any modifier keys.
?
Esc
h
f
a
m
v
c
r
q
t
u
o
,
.
/
s
n
e
;
Enter
[
\
k
i
l
=
-
0
′
1
2
3
4
5
6
7
8
9
→
↓
←
↑
Space
x
z
`
g
Apperently it’s more efficient to do it other way around, to compile programs into transformers, which are then useful as refecene and ground truth when analyzing “real” transformers.
See usage of TRACR in “Towards Automated Circuit Discovery for Mechanistic Interpretability” https://arxiv.org/pdf/2304.14997, for example.