INDEX

Explanations

what I thought

The neuron detects the hedge/quoting phrase “that’s what…” (as in “at least that’s what he claims,” “that’s what the data would show,” etc.).

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

 البته

-1.30

 even

-1.24

つい

-1.20

蹦

-1.20

 therefore

-1.16

 حتی

-1.16

ть

-1.14

だったが

-1.13

 идут

-1.12

why

-1.11

POSITIVE LOGITS

on

1.44

 linguri

1.32

';

1.28

 Klicken

1.25

uose

1.23

怍

1.22

':'

1.22

 fhould

1.21

ThreadPool

1.21

NIR

1.20

Activations Density 0.022%