INDEX

Explanations

the word "as" indicating comparisons or analogies

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

New Auto-Interp

Configuration

google/gemma-scope-9b-pt-res/layer_20/width_131k/average_l0_114

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

131,072

Data Type

float32

Hook Name

blocks.20.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

ResumeLayout

-0.56

 fubject

-0.53

vœ

-0.45

 løs

-0.44

 juſ

-0.43

 eneste

-0.43

räck

-0.43

etsk

-0.42

 juges

-0.42

rätt

-0.41

POSITIVE LOGITS

 CALLED

0.72

 called

0.63

 Called

0.63

called

0.63

称为

0.61

Called

0.60

叫做

0.57

ConstraintMaker

0.57

://

0.57

المناصب

0.57

Activations Density 0.010%

the word "as" indicating comparisons or analogies

No Comments

No Known Activations