INDEX

Explanations

reliable

The neuron fires on the word “reliable” (and its close variants) in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

⎆

-3.09

翃

-2.73

 quintessential

-2.72

 paltry

-2.70

actéristi

-2.66

 pivotal

-2.64

 wondrous

-2.61

 magnificent

-2.59

 astute

-2.58

讦

-2.58

POSITIVE LOGITS

3.55

just

3.25

𓆜

3.02

this

2.81

There

2.78

there

2.59

翾

2.56

its

2.56



2.52

It

2.44

Activations Density 0.010%