INDEX

Explanations

shared object files `.so`

The neuron is basically detecting the token “so” (as a discourse‐marker or filler).

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

 következő

-2.20

 Their

-2.20

 CALIFOR

-2.05

 There

-2.03

']

-2.00

 solcher

-1.99

 inhalt

-1.98

up

-1.97

啟動

-1.96

POSITIVE LOGITS

”

2.69

ꔹ

2.55

’

2.53

𖣔

2.52

so

2.50

 renowned

2.44

 stalwart

2.41

nienie

2.34

2.28

琉

2.27

Activations Density 0.005%