INDEX

Explanations

references to the pronoun "it."

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

 ostensibly

-0.50

 admittedly

-0.38

pleaños

-0.37

乍

-0.37

 ddelweddau

-0.37

 ovviamente

-0.37

 essentially

-0.36

 nonetheless

-0.36

 apparently

-0.35

ksikon

-0.35

POSITIVE LOGITS

 Worth

0.53

 worth

0.52

Accomplished

0.48

𓃵

0.48

oa̍t

0.48

 regarded

0.47

 quitted

0.46

 called

0.45

Worth

0.45

acord

0.44

Activations Density 0.051%

references to the pronoun "it."

No Comments

No Known Activations