INDEX

Explanations

instances of the word "triple" or similar variations indicating a quantity or structure involving three parts

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

hush

-0.47

Boyd

-0.46

Bord

-0.46

sians

-0.46

 DeSantis

-0.45

nav

-0.45

 μην

-0.44

oken

-0.44

 Boud

-0.44

bours

-0.44

POSITIVE LOGITS

 Triple

1.01

Triple

1.01

 triple

0.96

triple

0.89

IPLE

0.80

 tripled

0.70

 triples

0.69

 Whipple

0.64

 vectorielles

0.63

 Verſ

0.62

Activations Density 0.004%

instances of the word "triple" or similar variations indicating a quantity or structure involving three parts

No Comments

No Known Activations