INDEX
Explanations
cleanup/disaster response
news about oil-related incidents and environmental impacts.
The neuron fires on words describing post‐incident cleanup and recovery operations.
New Auto-Interp
Negative Logits
Mosque
-0.07
(light
-0.07
Intro
-0.07
Mode
-0.06
joke
-0.06
yerinde
-0.06
hour
-0.06
hides
-0.06
experience
-0.06
-Level
-0.06
POSITIVE LOGITS
ратно
0.06
ر
0.06
INFRINGEMENT
0.06
ิว
0.06
celestial
0.06
거
0.06
297
0.06
dlouh
0.06
rehears
0.06
404
0.06
Activations Density 0.020%