INDEX

Explanations

phrases indicating the frequency or commonality of occurrences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-9B-IT @ 20-gemmascope-res-131k

Configuration

google/gemma-scope-9b-it-res/layer_20/width_131k/average_l0_81

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

存于互联网档案馆

-0.56

parsedMessage

-0.53

 Wikimedijinoj

-0.47

 مشين

-0.47

 estekak

-0.47

Datuak

-0.46

󠁿

-0.45

 EventArgs

-0.45

exitRule

-0.44

chaun

-0.44

POSITIVE LOGITS

 externi

0.52

 linkovi

0.40

ecap

0.39

SharedCtor

0.38

LookAnd

0.37

 relâche

0.36

 nhung

0.36

Jîn

0.36

 erwartet

0.34

cipais

0.34

Activations Density 0.000%

phrases indicating the frequency or commonality of occurrences

No Comments

No Known Activations