INDEX

Explanations

let's break down

This neuron detects discourse-level transition or “call-to-action” phrases—i.e. sentence-initial cues like “So,” “let’s,” “start,” “here’s,” “basically,” etc., that shift or guide the flow of the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 ALWAYS

0.84

 Broadcasting

0.84

 TIPS

0.81

 EVERY

0.80

 spouses

0.80

 pleaded

0.79

 Gaur

0.78

Nxa

0.76

ANY

0.75

 Skating

0.73

POSITIVE LOGITS

vamos

0.97

早速

0.79

 bahas

0.77

Vamos

0.76

Lets

0.75

let

0.73

let

0.72

まず

0.72

まずは

0.71

dey

0.71

Activations Density 0.174%