INDEX

Explanations

HTML closing tags

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_33/width_131k/average_l0_91

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.33.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

RegressionTest

-0.59

 tartalomajánló

-0.59

ValueStyle

-0.53

ThroughAttribute

-0.51

MethodManager

-0.50

 الحره

-0.50

 AssemblyProduct

-0.50

SourceChecksum

-0.49

Personendaten

-0.48

FormTagHelper

-0.48

POSITIVE LOGITS

nav

0.47

 مرئيه

0.42

nav

0.41

 دام

0.38

textAppearance

0.34

 Injectable

0.33

NAV

0.32

Contributed

0.31

navigation

0.31

 Đình

0.31

Activations Density 0.011%

HTML closing tags

No Comments

No Known Activations