INDEX

Explanations

actions related to attempts and intentions to cause harm or disruption

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

Personendaten

-0.56

MessageOf

-0.52

########.

-0.52

 AssemblyCulture

-0.51

PMailer

-0.48

SharedCtor

-0.47

 GenerationType

-0.45

__":

-0.44

 bibliographies

-0.44

 tewas

-0.43

POSITIVE LOGITS

migrationBuilder

0.34

 wants

0.34

 Staates

0.33

dropIfExists

0.33

upassen

0.32

beginTransaction

0.32

/***/

0.32

pherals

0.32

lay

0.31

wikkel

0.31

Activations Density 0.033%

actions related to attempts and intentions to cause harm or disruption

No Comments

No Known Activations