INDEX

Explanations

get done

The neuron strongly activates on contiguous multi‐word content phrases (e.g. “hooked up,” “code running,” “get it working on,” “Angular JS working”), effectively marking spans of several successive content words.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

惆

-1.12

ilets

-1.10

 prevención

-1.03

 cóc

-1.01

//////////

-0.97

Mål

-0.94

 officio

-0.94

 candidats

-0.93

constructed

-0.93

 portatil

-0.92

POSITIVE LOGITS

 into

1.48

 involved

1.35

 started

1.28

 ASAP

1.19

out

1.11

↵↵↵↵↵↵↵↵

1.08

↵↵↵↵↵↵↵

1.08

rid

1.07

 issues

1.05

↵↵↵↵↵↵↵↵↵↵

1.02

Activations Density 0.065%