INDEX

Explanations

identifiers or codes, possibly related to database entries or classifications

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

 старости

-0.45

⤹

-0.40

OGND

-0.40

saraba

-0.39

 somehow

-0.39

せっかく

-0.39

üli

-0.39

 indé

-0.38

 magically

-0.37

 rapat

-0.36

POSITIVE LOGITS

 PeEnEo

0.57

DeleteBehavior

0.57

govina

0.47

keletal

0.44

 snippetHide

0.44

XMLSchema

0.43

acebook

0.43

TypedDataSet

0.42

 ſche

0.42

ConverterFactory

0.42

Activations Density 0.004%

identifiers or codes, possibly related to database entries or classifications

No Comments

No Known Activations