INDEX

Explanations

here or there

The neuron flags introductory locative/existential cues (e.g. “there,” “here,” “lo,” “in”) often used in inverted or expletive constructions (“there sits…,” “here’s…,” “in fell…,” “lo and behold…”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 semelh

-1.84

 Répondre

-1.73

橚

-1.72

 Trouver

-1.71

蔸

-1.69

 almoço

-1.69

 Chaque

-1.69

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

-1.67

Ꮴ

-1.64

鹇

-1.64

POSITIVE LOGITS

on

2.06

at

1.93

two

1.81

 without

1.66

in

1.59

 three

1.55

all

1.52

 four

1.51

TWO

1.44

 this

1.39

Activations Density 0.016%