INDEX

Explanations

Tahoe, heist, slowdown, Boltzmann, osha

The neuron fires on relatively long, low‐frequency tokens—especially proper names and specialized terms.

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

-3.09

-2.97

’

-2.88

-2.78

-2.75

鋨

-2.53

-2.45

-2.44

ープン

-2.42

ſſel

-2.41

POSITIVE LOGITS

 Because

2.73

 However

2.72

 Which

2.63

 Боль

2.34

 Since

2.30

of

2.27

 lounging

2.25

riever

2.23

ﾍ

2.17

We

2.14

Activations Density 0.002%